crates/common/src/models.rs

@@ -1412,7 +1412,7 @@ pub mod artifact {
         pub content_type: Option<String>,
         /// Size of the latest version's content in bytes
         pub size_bytes: Option<i64>,
-        /// Execution that produced this artifact (no FK — execution is a hypertable)
+        /// Execution that produced this artifact (no FK by design)
         pub execution: Option<Id>,
         /// Structured JSONB data for progress artifacts or metadata
         pub data: Option<serde_json::Value>,

crates/common/src/repositories/analytics.rs

@@ -80,7 +80,7 @@ pub struct EnforcementVolumeBucket {
     pub enforcement_count: i64,
 }
 
-/// A single hourly bucket of execution volume (from execution hypertable directly).
+/// A single hourly bucket of execution volume (from the execution table directly).
 #[derive(Debug, Clone, Serialize, FromRow)]
 pub struct ExecutionVolumeBucket {
     /// Start of the 1-hour bucket
@@ -468,7 +468,7 @@ impl AnalyticsRepository {
     }
 
     // =======================================================================
-    // Execution volume (from execution hypertable directly)
+    // Execution volume (from the execution table directly)
     // =======================================================================
 
     /// Query the `execution_volume_hourly` continuous aggregate for execution

crates/common/src/repositories/event.rs

@@ -65,6 +65,12 @@ pub struct EnforcementSearchResult {
     pub total: u64,
 }
 
+#[derive(Debug, Clone)]
+pub struct EnforcementCreateOrGetResult {
+    pub enforcement: Enforcement,
+    pub created: bool,
+}
+
 /// Repository for Event operations
 pub struct EventRepository;
 
@@ -493,11 +499,7 @@ impl EnforcementRepository {
         Ok(enforcement)
     }
 
-    /// Update an enforcement using the loaded row's hypertable keys.
-    ///
-    /// This avoids wide scans across compressed chunks by including both the
-    /// partitioning column (`created`) and compression segment key (`rule_ref`)
-    /// in the locator.
+    /// Update an enforcement using the loaded row's primary key.
     pub async fn update_loaded<'e, E>(
         executor: E,
         enforcement: &Enforcement,
@@ -510,19 +512,73 @@ impl EnforcementRepository {
             return Ok(enforcement.clone());
         }
 
-        let rule_ref = enforcement.rule_ref.clone();
-
         Self::update_with_locator(executor, input, |query| {
             query.push(" WHERE id = ");
             query.push_bind(enforcement.id);
-            query.push(" AND created = ");
-            query.push_bind(enforcement.created);
-            query.push(" AND rule_ref = ");
-            query.push_bind(rule_ref);
         })
         .await
     }
 
+    pub async fn update_loaded_if_status<'e, E>(
+        executor: E,
+        enforcement: &Enforcement,
+        expected_status: EnforcementStatus,
+        input: UpdateEnforcementInput,
+    ) -> Result<Option<Enforcement>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        if input.status.is_none() && input.payload.is_none() && input.resolved_at.is_none() {
+            return Ok(Some(enforcement.clone()));
+        }
+
+        let mut query = QueryBuilder::new("UPDATE enforcement SET ");
+        let mut has_updates = false;
+
+        if let Some(status) = input.status {
+            query.push("status = ");
+            query.push_bind(status);
+            has_updates = true;
+        }
+
+        if let Some(payload) = &input.payload {
+            if has_updates {
+                query.push(", ");
+            }
+            query.push("payload = ");
+            query.push_bind(payload);
+            has_updates = true;
+        }
+
+        if let Some(resolved_at) = input.resolved_at {
+            if has_updates {
+                query.push(", ");
+            }
+            query.push("resolved_at = ");
+            query.push_bind(resolved_at);
+            has_updates = true;
+        }
+
+        if !has_updates {
+            return Ok(Some(enforcement.clone()));
+        }
+
+        query.push(" WHERE id = ");
+        query.push_bind(enforcement.id);
+        query.push(" AND status = ");
+        query.push_bind(expected_status);
+        query.push(
+            " RETURNING id, rule, rule_ref, trigger_ref, config, event, status, payload, \
+             condition, conditions, created, resolved_at",
+        );
+
+        query
+            .build_query_as::<Enforcement>()
+            .fetch_optional(executor)
+            .await
+            .map_err(Into::into)
+    }
+
     /// Find enforcements by rule ID
     pub async fn find_by_rule<'e, E>(executor: E, rule_id: Id) -> Result<Vec<Enforcement>>
     where
@@ -589,6 +645,90 @@ impl EnforcementRepository {
         Ok(enforcements)
     }
 
+    pub async fn find_by_rule_and_event<'e, E>(
+        executor: E,
+        rule_id: Id,
+        event_id: Id,
+    ) -> Result<Option<Enforcement>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        sqlx::query_as::<_, Enforcement>(
+            r#"
+            SELECT id, rule, rule_ref, trigger_ref, config, event, status, payload,
+                   condition, conditions, created, resolved_at
+            FROM enforcement
+            WHERE rule = $1 AND event = $2
+            LIMIT 1
+            "#,
+        )
+        .bind(rule_id)
+        .bind(event_id)
+        .fetch_optional(executor)
+        .await
+        .map_err(Into::into)
+    }
+
+    pub async fn create_or_get_by_rule_event<'e, E>(
+        executor: E,
+        input: CreateEnforcementInput,
+    ) -> Result<EnforcementCreateOrGetResult>
+    where
+        E: Executor<'e, Database = Postgres> + Copy + 'e,
+    {
+        let (Some(rule_id), Some(event_id)) = (input.rule, input.event) else {
+            let enforcement = Self::create(executor, input).await?;
+            return Ok(EnforcementCreateOrGetResult {
+                enforcement,
+                created: true,
+            });
+        };
+
+        let inserted = sqlx::query_as::<_, Enforcement>(
+            r#"
+            INSERT INTO enforcement (rule, rule_ref, trigger_ref, config, event, status,
+                                     payload, condition, conditions)
+            VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9)
+            ON CONFLICT (rule, event) WHERE rule IS NOT NULL AND event IS NOT NULL DO NOTHING
+            RETURNING id, rule, rule_ref, trigger_ref, config, event, status, payload,
+                      condition, conditions, created, resolved_at
+            "#,
+        )
+        .bind(input.rule)
+        .bind(&input.rule_ref)
+        .bind(&input.trigger_ref)
+        .bind(&input.config)
+        .bind(input.event)
+        .bind(input.status)
+        .bind(&input.payload)
+        .bind(input.condition)
+        .bind(&input.conditions)
+        .fetch_optional(executor)
+        .await?;
+
+        if let Some(enforcement) = inserted {
+            return Ok(EnforcementCreateOrGetResult {
+                enforcement,
+                created: true,
+            });
+        }
+
+        let enforcement = Self::find_by_rule_and_event(executor, rule_id, event_id)
+            .await?
+            .ok_or_else(|| {
+                anyhow::anyhow!(
+                    "enforcement for rule {} and event {} disappeared after dedupe conflict",
+                    rule_id,
+                    event_id
+                )
+            })?;
+
+        Ok(EnforcementCreateOrGetResult {
+            enforcement,
+            created: false,
+        })
+    }
+
     /// Search enforcements with all filters pushed into SQL.
     ///
     /// All filter fields are combinable (AND). Pagination is server-side.

crates/common/src/repositories/execution.rs

@@ -4,7 +4,8 @@ use chrono::{DateTime, Utc};
 
 use crate::models::{enums::ExecutionStatus, execution::*, Id, JsonDict};
 use crate::Result;
-use sqlx::{Executor, Postgres, QueryBuilder};
+use sqlx::{Executor, PgConnection, PgPool, Postgres, QueryBuilder};
+use tokio::time::{sleep, Duration};
 
 use super::{Create, Delete, FindById, List, Repository, Update};
 
@@ -41,6 +42,18 @@ pub struct ExecutionSearchResult {
     pub total: u64,
 }
 
+#[derive(Debug, Clone)]
+pub struct WorkflowTaskExecutionCreateOrGetResult {
+    pub execution: Execution,
+    pub created: bool,
+}
+
+#[derive(Debug, Clone)]
+pub struct EnforcementExecutionCreateOrGetResult {
+    pub execution: Execution,
+    pub created: bool,
+}
+
 /// An execution row with optional `rule_ref` / `trigger_ref` populated from
 /// the joined `enforcement` table. This avoids a separate in-memory lookup.
 #[derive(Debug, Clone, sqlx::FromRow)]
@@ -209,6 +222,550 @@ impl Update for ExecutionRepository {
 }
 
 impl ExecutionRepository {
+    pub async fn find_top_level_by_enforcement<'e, E>(
+        executor: E,
+        enforcement_id: Id,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        let sql = format!(
+            "SELECT {SELECT_COLUMNS} \
+             FROM execution \
+             WHERE enforcement = $1
+               AND parent IS NULL
+               AND (config IS NULL OR NOT (config ? 'retry_of')) \
+             ORDER BY created ASC \
+             LIMIT 1"
+        );
+
+        sqlx::query_as::<_, Execution>(&sql)
+            .bind(enforcement_id)
+            .fetch_optional(executor)
+            .await
+            .map_err(Into::into)
+    }
+
+    pub async fn create_top_level_for_enforcement_if_absent<'e, E>(
+        executor: E,
+        input: CreateExecutionInput,
+        enforcement_id: Id,
+    ) -> Result<EnforcementExecutionCreateOrGetResult>
+    where
+        E: Executor<'e, Database = Postgres> + Copy + 'e,
+    {
+        let inserted = sqlx::query_as::<_, Execution>(&format!(
+            "INSERT INTO execution \
+             (action, action_ref, config, env_vars, parent, enforcement, executor, worker, status, result, workflow_task) \
+             VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11) \
+             ON CONFLICT (enforcement)
+             WHERE enforcement IS NOT NULL
+               AND parent IS NULL
+               AND (config IS NULL OR NOT (config ? 'retry_of'))
+             DO NOTHING \
+             RETURNING {SELECT_COLUMNS}"
+        ))
+        .bind(input.action)
+        .bind(&input.action_ref)
+        .bind(&input.config)
+        .bind(&input.env_vars)
+        .bind(input.parent)
+        .bind(input.enforcement)
+        .bind(input.executor)
+        .bind(input.worker)
+        .bind(input.status)
+        .bind(&input.result)
+        .bind(sqlx::types::Json(&input.workflow_task))
+        .fetch_optional(executor)
+        .await?;
+
+        if let Some(execution) = inserted {
+            return Ok(EnforcementExecutionCreateOrGetResult {
+                execution,
+                created: true,
+            });
+        }
+
+        let execution = Self::find_top_level_by_enforcement(executor, enforcement_id)
+            .await?
+            .ok_or_else(|| {
+                anyhow::anyhow!(
+                    "top-level execution for enforcement {} disappeared after dedupe conflict",
+                    enforcement_id
+                )
+            })?;
+
+        Ok(EnforcementExecutionCreateOrGetResult {
+            execution,
+            created: false,
+        })
+    }
+
+    async fn claim_workflow_task_dispatch<'e, E>(
+        executor: E,
+        workflow_execution_id: Id,
+        task_name: &str,
+        task_index: Option<i32>,
+    ) -> Result<bool>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        let inserted: Option<(i64,)> = sqlx::query_as(
+            "INSERT INTO workflow_task_dispatch (workflow_execution, task_name, task_index)
+             VALUES ($1, $2, $3)
+             ON CONFLICT (workflow_execution, task_name, COALESCE(task_index, -1)) DO NOTHING
+             RETURNING id",
+        )
+        .bind(workflow_execution_id)
+        .bind(task_name)
+        .bind(task_index)
+        .fetch_optional(executor)
+        .await?;
+
+        Ok(inserted.is_some())
+    }
+
+    async fn assign_workflow_task_dispatch_execution<'e, E>(
+        executor: E,
+        workflow_execution_id: Id,
+        task_name: &str,
+        task_index: Option<i32>,
+        execution_id: Id,
+    ) -> Result<()>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        sqlx::query(
+            "UPDATE workflow_task_dispatch
+             SET execution_id = COALESCE(execution_id, $4)
+             WHERE workflow_execution = $1
+               AND task_name = $2
+               AND task_index IS NOT DISTINCT FROM $3",
+        )
+        .bind(workflow_execution_id)
+        .bind(task_name)
+        .bind(task_index)
+        .bind(execution_id)
+        .execute(executor)
+        .await?;
+
+        Ok(())
+    }
+
+    async fn lock_workflow_task_dispatch<'e, E>(
+        executor: E,
+        workflow_execution_id: Id,
+        task_name: &str,
+        task_index: Option<i32>,
+    ) -> Result<Option<Option<Id>>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        let row: Option<(Option<i64>,)> = sqlx::query_as(
+            "SELECT execution_id
+             FROM workflow_task_dispatch
+             WHERE workflow_execution = $1
+               AND task_name = $2
+               AND task_index IS NOT DISTINCT FROM $3
+             FOR UPDATE",
+        )
+        .bind(workflow_execution_id)
+        .bind(task_name)
+        .bind(task_index)
+        .fetch_optional(executor)
+        .await?;
+
+        // Map the outer Option to distinguish three cases:
+        // - None            → no row exists
+        // - Some(None)      → row exists but execution_id is still NULL (mid-creation)
+        // - Some(Some(id))  → row exists with a completed execution_id
+        Ok(row.map(|(execution_id,)| execution_id))
+    }
+
+    async fn create_workflow_task_if_absent_in_conn(
+        conn: &mut PgConnection,
+        input: CreateExecutionInput,
+        workflow_execution_id: Id,
+        task_name: &str,
+        task_index: Option<i32>,
+    ) -> Result<WorkflowTaskExecutionCreateOrGetResult> {
+        let claimed = Self::claim_workflow_task_dispatch(
+            &mut *conn,
+            workflow_execution_id,
+            task_name,
+            task_index,
+        )
+        .await?;
+
+        if claimed {
+            let execution = Self::create(&mut *conn, input).await?;
+            Self::assign_workflow_task_dispatch_execution(
+                &mut *conn,
+                workflow_execution_id,
+                task_name,
+                task_index,
+                execution.id,
+            )
+            .await?;
+
+            return Ok(WorkflowTaskExecutionCreateOrGetResult {
+                execution,
+                created: true,
+            });
+        }
+
+        let dispatch_state = Self::lock_workflow_task_dispatch(
+            &mut *conn,
+            workflow_execution_id,
+            task_name,
+            task_index,
+        )
+        .await?;
+
+        match dispatch_state {
+            Some(Some(existing_execution_id)) => {
+                // Row exists with execution_id — return the existing execution.
+                let execution = Self::find_by_id(&mut *conn, existing_execution_id)
+                    .await?
+                    .ok_or_else(|| {
+                        anyhow::anyhow!(
+                            "workflow child execution {} missing for workflow_execution {} task '{}' index {:?}",
+                            existing_execution_id,
+                            workflow_execution_id,
+                            task_name,
+                            task_index
+                        )
+                    })?;
+
+                Ok(WorkflowTaskExecutionCreateOrGetResult {
+                    execution,
+                    created: false,
+                })
+            }
+
+            Some(None) => {
+                // Row exists but execution_id is still NULL: another transaction is
+                // mid-creation (between claim and assign). Retry until it's filled in.
+                // If the original creator's transaction rolled back, the row also
+                // disappears — handled by the `None` branch inside the loop.
+                'wait: {
+                    for _ in 0..20_u32 {
+                        sleep(Duration::from_millis(50)).await;
+                        match Self::lock_workflow_task_dispatch(
+                            &mut *conn,
+                            workflow_execution_id,
+                            task_name,
+                            task_index,
+                        )
+                        .await?
+                        {
+                            Some(Some(execution_id)) => {
+                                let execution =
+                                    Self::find_by_id(&mut *conn, execution_id).await?.ok_or_else(
+                                        || {
+                                            anyhow::anyhow!(
+                                                "workflow child execution {} missing for workflow_execution {} task '{}' index {:?}",
+                                                execution_id,
+                                                workflow_execution_id,
+                                                task_name,
+                                                task_index
+                                            )
+                                        },
+                                    )?;
+                                return Ok(WorkflowTaskExecutionCreateOrGetResult {
+                                    execution,
+                                    created: false,
+                                });
+                            }
+                            Some(None) => {}     // still NULL, keep waiting
+                            None => break 'wait, // row rolled back; fall through to re-claim
+                        }
+                    }
+                    // Exhausted all retries without the execution_id being set.
+                    return Err(anyhow::anyhow!(
+                        "Timed out waiting for workflow task dispatch execution_id to be set \
+                         for workflow_execution {} task '{}' index {:?}",
+                        workflow_execution_id,
+                        task_name,
+                        task_index
+                    )
+                    .into());
+                }
+
+                // Row disappeared (original creator rolled back) — re-claim and create.
+                let re_claimed = Self::claim_workflow_task_dispatch(
+                    &mut *conn,
+                    workflow_execution_id,
+                    task_name,
+                    task_index,
+                )
+                .await?;
+                if !re_claimed {
+                    return Err(anyhow::anyhow!(
+                        "Workflow task dispatch for workflow_execution {} task '{}' index {:?} \
+                         was reclaimed by another executor after rollback",
+                        workflow_execution_id,
+                        task_name,
+                        task_index
+                    )
+                    .into());
+                }
+                let execution = Self::create(&mut *conn, input).await?;
+                Self::assign_workflow_task_dispatch_execution(
+                    &mut *conn,
+                    workflow_execution_id,
+                    task_name,
+                    task_index,
+                    execution.id,
+                )
+                .await?;
+                Ok(WorkflowTaskExecutionCreateOrGetResult {
+                    execution,
+                    created: true,
+                })
+            }
+
+            None => {
+                // No row at all — the original INSERT was rolled back before we arrived.
+                // Attempt to re-claim and create as if this were a fresh dispatch.
+                let re_claimed = Self::claim_workflow_task_dispatch(
+                    &mut *conn,
+                    workflow_execution_id,
+                    task_name,
+                    task_index,
+                )
+                .await?;
+                if !re_claimed {
+                    return Err(anyhow::anyhow!(
+                        "Workflow task dispatch for workflow_execution {} task '{}' index {:?} \
+                         was claimed by another executor",
+                        workflow_execution_id,
+                        task_name,
+                        task_index
+                    )
+                    .into());
+                }
+                let execution = Self::create(&mut *conn, input).await?;
+                Self::assign_workflow_task_dispatch_execution(
+                    &mut *conn,
+                    workflow_execution_id,
+                    task_name,
+                    task_index,
+                    execution.id,
+                )
+                .await?;
+                Ok(WorkflowTaskExecutionCreateOrGetResult {
+                    execution,
+                    created: true,
+                })
+            }
+        }
+    }
+
+    pub async fn create_workflow_task_if_absent(
+        pool: &PgPool,
+        input: CreateExecutionInput,
+        workflow_execution_id: Id,
+        task_name: &str,
+        task_index: Option<i32>,
+    ) -> Result<WorkflowTaskExecutionCreateOrGetResult> {
+        let mut conn = pool.acquire().await?;
+        sqlx::query("BEGIN").execute(&mut *conn).await?;
+
+        let result = Self::create_workflow_task_if_absent_in_conn(
+            &mut conn,
+            input,
+            workflow_execution_id,
+            task_name,
+            task_index,
+        )
+        .await;
+
+        match result {
+            Ok(result) => {
+                sqlx::query("COMMIT").execute(&mut *conn).await?;
+                Ok(result)
+            }
+            Err(err) => {
+                sqlx::query("ROLLBACK").execute(&mut *conn).await?;
+                Err(err)
+            }
+        }
+    }
+
+    pub async fn create_workflow_task_if_absent_with_conn(
+        conn: &mut PgConnection,
+        input: CreateExecutionInput,
+        workflow_execution_id: Id,
+        task_name: &str,
+        task_index: Option<i32>,
+    ) -> Result<WorkflowTaskExecutionCreateOrGetResult> {
+        Self::create_workflow_task_if_absent_in_conn(
+            conn,
+            input,
+            workflow_execution_id,
+            task_name,
+            task_index,
+        )
+        .await
+    }
+
+    pub async fn claim_for_scheduling<'e, E>(
+        executor: E,
+        id: Id,
+        claiming_executor: Option<Id>,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        let sql = format!(
+            "UPDATE execution \
+             SET status = $2, executor = COALESCE($3, executor), updated = NOW() \
+             WHERE id = $1 AND status = $4 \
+             RETURNING {SELECT_COLUMNS}"
+        );
+
+        sqlx::query_as::<_, Execution>(&sql)
+            .bind(id)
+            .bind(ExecutionStatus::Scheduling)
+            .bind(claiming_executor)
+            .bind(ExecutionStatus::Requested)
+            .fetch_optional(executor)
+            .await
+            .map_err(Into::into)
+    }
+
+    pub async fn reclaim_stale_scheduling<'e, E>(
+        executor: E,
+        id: Id,
+        claiming_executor: Option<Id>,
+        stale_before: DateTime<Utc>,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        let sql = format!(
+            "UPDATE execution \
+             SET executor = COALESCE($2, executor), updated = NOW() \
+             WHERE id = $1 AND status = $3 AND updated <= $4 \
+             RETURNING {SELECT_COLUMNS}"
+        );
+
+        sqlx::query_as::<_, Execution>(&sql)
+            .bind(id)
+            .bind(claiming_executor)
+            .bind(ExecutionStatus::Scheduling)
+            .bind(stale_before)
+            .fetch_optional(executor)
+            .await
+            .map_err(Into::into)
+    }
+
+    pub async fn update_if_status<'e, E>(
+        executor: E,
+        id: Id,
+        expected_status: ExecutionStatus,
+        input: UpdateExecutionInput,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        if input.status.is_none()
+            && input.result.is_none()
+            && input.executor.is_none()
+            && input.worker.is_none()
+            && input.started_at.is_none()
+            && input.workflow_task.is_none()
+        {
+            return Self::find_by_id(executor, id).await;
+        }
+
+        Self::update_with_locator_optional(executor, input, |query| {
+            query.push(" WHERE id = ").push_bind(id);
+            query.push(" AND status = ").push_bind(expected_status);
+        })
+        .await
+    }
+
+    pub async fn update_if_status_and_updated_before<'e, E>(
+        executor: E,
+        id: Id,
+        expected_status: ExecutionStatus,
+        stale_before: DateTime<Utc>,
+        input: UpdateExecutionInput,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        if input.status.is_none()
+            && input.result.is_none()
+            && input.executor.is_none()
+            && input.worker.is_none()
+            && input.started_at.is_none()
+            && input.workflow_task.is_none()
+        {
+            return Self::find_by_id(executor, id).await;
+        }
+
+        Self::update_with_locator_optional(executor, input, |query| {
+            query.push(" WHERE id = ").push_bind(id);
+            query.push(" AND status = ").push_bind(expected_status);
+            query.push(" AND updated < ").push_bind(stale_before);
+        })
+        .await
+    }
+
+    pub async fn update_if_status_and_updated_at<'e, E>(
+        executor: E,
+        id: Id,
+        expected_status: ExecutionStatus,
+        expected_updated: DateTime<Utc>,
+        input: UpdateExecutionInput,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        if input.status.is_none()
+            && input.result.is_none()
+            && input.executor.is_none()
+            && input.worker.is_none()
+            && input.started_at.is_none()
+            && input.workflow_task.is_none()
+        {
+            return Self::find_by_id(executor, id).await;
+        }
+
+        Self::update_with_locator_optional(executor, input, |query| {
+            query.push(" WHERE id = ").push_bind(id);
+            query.push(" AND status = ").push_bind(expected_status);
+            query.push(" AND updated = ").push_bind(expected_updated);
+        })
+        .await
+    }
+
+    pub async fn revert_scheduled_to_requested<'e, E>(
+        executor: E,
+        id: Id,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        let sql = format!(
+            "UPDATE execution \
+             SET status = $2, worker = NULL, executor = NULL, updated = NOW() \
+             WHERE id = $1 AND status = $3 \
+             RETURNING {SELECT_COLUMNS}"
+        );
+
+        sqlx::query_as::<_, Execution>(&sql)
+            .bind(id)
+            .bind(ExecutionStatus::Requested)
+            .bind(ExecutionStatus::Scheduled)
+            .fetch_optional(executor)
+            .await
+            .map_err(Into::into)
+    }
+
     async fn update_with_locator<'e, E, F>(
         executor: E,
         input: UpdateExecutionInput,
@@ -274,10 +831,72 @@ impl ExecutionRepository {
             .map_err(Into::into)
     }
 
-    /// Update an execution using the loaded row's hypertable keys.
-    ///
-    /// Including both the partition key (`created`) and compression segment key
-    /// (`action_ref`) avoids broad scans across compressed chunks.
+    async fn update_with_locator_optional<'e, E, F>(
+        executor: E,
+        input: UpdateExecutionInput,
+        where_clause: F,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+        F: FnOnce(&mut QueryBuilder<'_, Postgres>),
+    {
+        let mut query = QueryBuilder::new("UPDATE execution SET ");
+        let mut has_updates = false;
+
+        if let Some(status) = input.status {
+            query.push("status = ").push_bind(status);
+            has_updates = true;
+        }
+        if let Some(result) = &input.result {
+            if has_updates {
+                query.push(", ");
+            }
+            query.push("result = ").push_bind(result);
+            has_updates = true;
+        }
+        if let Some(executor_id) = input.executor {
+            if has_updates {
+                query.push(", ");
+            }
+            query.push("executor = ").push_bind(executor_id);
+            has_updates = true;
+        }
+        if let Some(worker_id) = input.worker {
+            if has_updates {
+                query.push(", ");
+            }
+            query.push("worker = ").push_bind(worker_id);
+            has_updates = true;
+        }
+        if let Some(started_at) = input.started_at {
+            if has_updates {
+                query.push(", ");
+            }
+            query.push("started_at = ").push_bind(started_at);
+            has_updates = true;
+        }
+        if let Some(workflow_task) = &input.workflow_task {
+            if has_updates {
+                query.push(", ");
+            }
+            query
+                .push("workflow_task = ")
+                .push_bind(sqlx::types::Json(workflow_task));
+        }
+
+        query.push(", updated = NOW()");
+        where_clause(&mut query);
+        query.push(" RETURNING ");
+        query.push(SELECT_COLUMNS);
+
+        query
+            .build_query_as::<Execution>()
+            .fetch_optional(executor)
+            .await
+            .map_err(Into::into)
+    }
+
+    /// Update an execution using the loaded row's primary key.
     pub async fn update_loaded<'e, E>(
         executor: E,
         execution: &Execution,
@@ -296,12 +915,8 @@ impl ExecutionRepository {
             return Ok(execution.clone());
         }
 
-        let action_ref = execution.action_ref.clone();
-
         Self::update_with_locator(executor, input, |query| {
             query.push(" WHERE id = ").push_bind(execution.id);
-            query.push(" AND created = ").push_bind(execution.created);
-            query.push(" AND action_ref = ").push_bind(action_ref);
         })
         .await
     }
@@ -356,6 +971,34 @@ impl ExecutionRepository {
             .map_err(Into::into)
     }
 
+    pub async fn find_by_workflow_task<'e, E>(
+        executor: E,
+        workflow_execution_id: Id,
+        task_name: &str,
+        task_index: Option<i32>,
+    ) -> Result<Option<Execution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        let sql = format!(
+            "SELECT {SELECT_COLUMNS} \
+             FROM execution \
+             WHERE workflow_task->>'workflow_execution' = $1::text \
+               AND workflow_task->>'task_name' = $2 \
+               AND (workflow_task->>'task_index')::int IS NOT DISTINCT FROM $3 \
+             ORDER BY created ASC \
+             LIMIT 1"
+        );
+
+        sqlx::query_as::<_, Execution>(&sql)
+            .bind(workflow_execution_id.to_string())
+            .bind(task_name)
+            .bind(task_index)
+            .fetch_optional(executor)
+            .await
+            .map_err(Into::into)
+    }
+
     /// Find all child executions for a given parent execution ID.
     ///
     /// Returns child executions ordered by creation time (ascending),

crates/common/src/repositories/execution_admission.rs

@@ -0,0 +1,909 @@
+use chrono::{DateTime, Utc};
+use sqlx::{PgPool, Postgres, Row, Transaction};
+
+use crate::error::Result;
+use crate::models::Id;
+use crate::repositories::queue_stats::{QueueStatsRepository, UpsertQueueStatsInput};
+
+#[derive(Debug, Clone)]
+pub struct AdmissionSlotAcquireOutcome {
+    pub acquired: bool,
+    pub current_count: u32,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum AdmissionEnqueueOutcome {
+    Acquired,
+    Enqueued,
+}
+
+#[derive(Debug, Clone)]
+pub struct AdmissionSlotReleaseOutcome {
+    pub action_id: Id,
+    pub group_key: Option<String>,
+    pub next_execution_id: Option<Id>,
+}
+
+#[derive(Debug, Clone)]
+pub struct AdmissionQueuedRemovalOutcome {
+    pub action_id: Id,
+    pub group_key: Option<String>,
+    pub next_execution_id: Option<Id>,
+    pub execution_id: Id,
+    pub queue_order: i64,
+    pub enqueued_at: DateTime<Utc>,
+    pub removed_index: usize,
+}
+
+#[derive(Debug, Clone)]
+pub struct AdmissionQueueStats {
+    pub action_id: Id,
+    pub queue_length: usize,
+    pub active_count: u32,
+    pub max_concurrent: u32,
+    pub oldest_enqueued_at: Option<DateTime<Utc>>,
+    pub total_enqueued: u64,
+    pub total_completed: u64,
+}
+
+#[derive(Debug, Clone)]
+struct AdmissionState {
+    id: Id,
+    action_id: Id,
+    group_key: Option<String>,
+    max_concurrent: i32,
+}
+
+#[derive(Debug, Clone)]
+struct ExecutionEntry {
+    state_id: Id,
+    action_id: Id,
+    group_key: Option<String>,
+    status: String,
+    queue_order: i64,
+    enqueued_at: DateTime<Utc>,
+}
+
+pub struct ExecutionAdmissionRepository;
+
+impl ExecutionAdmissionRepository {
+    pub async fn enqueue(
+        pool: &PgPool,
+        max_queue_length: usize,
+        action_id: Id,
+        execution_id: Id,
+        max_concurrent: u32,
+        group_key: Option<String>,
+    ) -> Result<AdmissionEnqueueOutcome> {
+        let mut tx = pool.begin().await?;
+        let state = Self::lock_state(&mut tx, action_id, group_key, max_concurrent).await?;
+        let outcome =
+            Self::enqueue_in_state(&mut tx, &state, max_queue_length, execution_id, true).await?;
+        Self::refresh_queue_stats(&mut tx, action_id).await?;
+        tx.commit().await?;
+        Ok(outcome)
+    }
+
+    pub async fn wait_status(pool: &PgPool, execution_id: Id) -> Result<Option<bool>> {
+        let row = sqlx::query_scalar::<Postgres, bool>(
+            r#"
+            SELECT status = 'active'
+            FROM execution_admission_entry
+            WHERE execution_id = $1
+            "#,
+        )
+        .bind(execution_id)
+        .fetch_optional(pool)
+        .await?;
+
+        Ok(row)
+    }
+
+    pub async fn try_acquire(
+        pool: &PgPool,
+        action_id: Id,
+        execution_id: Id,
+        max_concurrent: u32,
+        group_key: Option<String>,
+    ) -> Result<AdmissionSlotAcquireOutcome> {
+        let mut tx = pool.begin().await?;
+        let state = Self::lock_state(&mut tx, action_id, group_key, max_concurrent).await?;
+        let active_count = Self::active_count(&mut tx, state.id).await? as u32;
+
+        let outcome = match Self::find_execution_entry(&mut tx, execution_id).await? {
+            Some(entry) if entry.status == "active" => AdmissionSlotAcquireOutcome {
+                acquired: true,
+                current_count: active_count,
+            },
+            Some(entry) if entry.status == "queued" && entry.state_id == state.id => {
+                let promoted =
+                    Self::maybe_promote_existing_queued(&mut tx, &state, execution_id).await?;
+                AdmissionSlotAcquireOutcome {
+                    acquired: promoted,
+                    current_count: active_count,
+                }
+            }
+            Some(_) => AdmissionSlotAcquireOutcome {
+                acquired: false,
+                current_count: active_count,
+            },
+            None => {
+                if active_count < max_concurrent
+                    && Self::queued_count(&mut tx, state.id).await? == 0
+                {
+                    let queue_order = Self::allocate_queue_order(&mut tx, state.id).await?;
+                    Self::insert_entry(
+                        &mut tx,
+                        state.id,
+                        execution_id,
+                        "active",
+                        queue_order,
+                        Utc::now(),
+                    )
+                    .await?;
+                    Self::increment_total_enqueued(&mut tx, state.id).await?;
+                    Self::refresh_queue_stats(&mut tx, action_id).await?;
+                    AdmissionSlotAcquireOutcome {
+                        acquired: true,
+                        current_count: active_count,
+                    }
+                } else {
+                    AdmissionSlotAcquireOutcome {
+                        acquired: false,
+                        current_count: active_count,
+                    }
+                }
+            }
+        };
+
+        tx.commit().await?;
+        Ok(outcome)
+    }
+
+    pub async fn release_active_slot(
+        pool: &PgPool,
+        execution_id: Id,
+    ) -> Result<Option<AdmissionSlotReleaseOutcome>> {
+        let mut tx = pool.begin().await?;
+        let Some(entry) = Self::find_execution_entry_for_update(&mut tx, execution_id).await?
+        else {
+            tx.commit().await?;
+            return Ok(None);
+        };
+
+        if entry.status != "active" {
+            tx.commit().await?;
+            return Ok(None);
+        }
+
+        let state = Self::lock_existing_state(&mut tx, entry.action_id, entry.group_key.clone())
+            .await?
+            .ok_or_else(|| {
+                crate::Error::internal("missing execution_admission_state for active execution")
+            })?;
+
+        sqlx::query("DELETE FROM execution_admission_entry WHERE execution_id = $1")
+            .bind(execution_id)
+            .execute(&mut *tx)
+            .await?;
+
+        Self::increment_total_completed(&mut tx, state.id).await?;
+
+        let next_execution_id = Self::promote_next_queued(&mut tx, &state).await?;
+        Self::refresh_queue_stats(&mut tx, state.action_id).await?;
+        tx.commit().await?;
+
+        Ok(Some(AdmissionSlotReleaseOutcome {
+            action_id: state.action_id,
+            group_key: state.group_key,
+            next_execution_id,
+        }))
+    }
+
+    pub async fn restore_active_slot(
+        pool: &PgPool,
+        execution_id: Id,
+        outcome: &AdmissionSlotReleaseOutcome,
+    ) -> Result<()> {
+        let mut tx = pool.begin().await?;
+        let state =
+            Self::lock_existing_state(&mut tx, outcome.action_id, outcome.group_key.clone())
+                .await?
+                .ok_or_else(|| {
+                    crate::Error::internal("missing execution_admission_state on restore")
+                })?;
+
+        if let Some(next_execution_id) = outcome.next_execution_id {
+            sqlx::query(
+                r#"
+                UPDATE execution_admission_entry
+                SET status = 'queued', activated_at = NULL
+                WHERE execution_id = $1
+                  AND state_id = $2
+                  AND status = 'active'
+                "#,
+            )
+            .bind(next_execution_id)
+            .bind(state.id)
+            .execute(&mut *tx)
+            .await?;
+        }
+
+        sqlx::query(
+            r#"
+            INSERT INTO execution_admission_entry (
+                state_id, execution_id, status, queue_order, enqueued_at, activated_at
+            ) VALUES ($1, $2, 'active', $3, NOW(), NOW())
+            ON CONFLICT (execution_id) DO UPDATE
+            SET state_id = EXCLUDED.state_id,
+                status = 'active',
+                activated_at = EXCLUDED.activated_at
+            "#,
+        )
+        .bind(state.id)
+        .bind(execution_id)
+        .bind(Self::allocate_queue_order(&mut tx, state.id).await?)
+        .execute(&mut *tx)
+        .await?;
+
+        sqlx::query(
+            r#"
+            UPDATE execution_admission_state
+            SET total_completed = GREATEST(total_completed - 1, 0)
+            WHERE id = $1
+            "#,
+        )
+        .bind(state.id)
+        .execute(&mut *tx)
+        .await?;
+
+        Self::refresh_queue_stats(&mut tx, state.action_id).await?;
+        tx.commit().await?;
+        Ok(())
+    }
+
+    pub async fn remove_queued_execution(
+        pool: &PgPool,
+        execution_id: Id,
+    ) -> Result<Option<AdmissionQueuedRemovalOutcome>> {
+        let mut tx = pool.begin().await?;
+        let Some(entry) = Self::find_execution_entry_for_update(&mut tx, execution_id).await?
+        else {
+            tx.commit().await?;
+            return Ok(None);
+        };
+
+        if entry.status != "queued" {
+            tx.commit().await?;
+            return Ok(None);
+        }
+
+        let state = Self::lock_existing_state(&mut tx, entry.action_id, entry.group_key.clone())
+            .await?
+            .ok_or_else(|| {
+                crate::Error::internal("missing execution_admission_state for queued execution")
+            })?;
+
+        let removed_index = sqlx::query_scalar::<Postgres, i64>(
+            r#"
+            SELECT COUNT(*)
+            FROM execution_admission_entry
+            WHERE state_id = $1
+              AND status = 'queued'
+              AND (enqueued_at, id) < (
+                    SELECT enqueued_at, id
+                    FROM execution_admission_entry
+                    WHERE execution_id = $2
+                )
+            "#,
+        )
+        .bind(state.id)
+        .bind(execution_id)
+        .fetch_one(&mut *tx)
+        .await? as usize;
+
+        sqlx::query("DELETE FROM execution_admission_entry WHERE execution_id = $1")
+            .bind(execution_id)
+            .execute(&mut *tx)
+            .await?;
+
+        let next_execution_id =
+            if Self::active_count(&mut tx, state.id).await? < state.max_concurrent as i64 {
+                Self::promote_next_queued(&mut tx, &state).await?
+            } else {
+                None
+            };
+
+        Self::refresh_queue_stats(&mut tx, state.action_id).await?;
+        tx.commit().await?;
+
+        Ok(Some(AdmissionQueuedRemovalOutcome {
+            action_id: state.action_id,
+            group_key: state.group_key,
+            next_execution_id,
+            execution_id,
+            queue_order: entry.queue_order,
+            enqueued_at: entry.enqueued_at,
+            removed_index,
+        }))
+    }
+
+    pub async fn restore_queued_execution(
+        pool: &PgPool,
+        outcome: &AdmissionQueuedRemovalOutcome,
+    ) -> Result<()> {
+        let mut tx = pool.begin().await?;
+        let state =
+            Self::lock_existing_state(&mut tx, outcome.action_id, outcome.group_key.clone())
+                .await?
+                .ok_or_else(|| {
+                    crate::Error::internal("missing execution_admission_state on queued restore")
+                })?;
+
+        if let Some(next_execution_id) = outcome.next_execution_id {
+            sqlx::query(
+                r#"
+                UPDATE execution_admission_entry
+                SET status = 'queued', activated_at = NULL
+                WHERE execution_id = $1
+                  AND state_id = $2
+                  AND status = 'active'
+                "#,
+            )
+            .bind(next_execution_id)
+            .bind(state.id)
+            .execute(&mut *tx)
+            .await?;
+        }
+
+        sqlx::query(
+            r#"
+            INSERT INTO execution_admission_entry (
+                state_id, execution_id, status, queue_order, enqueued_at, activated_at
+            ) VALUES ($1, $2, 'queued', $3, $4, NULL)
+            ON CONFLICT (execution_id) DO NOTHING
+            "#,
+        )
+        .bind(state.id)
+        .bind(outcome.execution_id)
+        .bind(outcome.queue_order)
+        .bind(outcome.enqueued_at)
+        .execute(&mut *tx)
+        .await?;
+
+        Self::refresh_queue_stats(&mut tx, state.action_id).await?;
+        tx.commit().await?;
+        Ok(())
+    }
+
+    pub async fn get_queue_stats(
+        pool: &PgPool,
+        action_id: Id,
+    ) -> Result<Option<AdmissionQueueStats>> {
+        let row = sqlx::query(
+            r#"
+            WITH state_rows AS (
+                SELECT
+                    COUNT(*) AS state_count,
+                    COALESCE(SUM(max_concurrent), 0) AS max_concurrent,
+                    COALESCE(SUM(total_enqueued), 0) AS total_enqueued,
+                    COALESCE(SUM(total_completed), 0) AS total_completed
+                FROM execution_admission_state
+                WHERE action_id = $1
+            ),
+            entry_rows AS (
+                SELECT
+                    COUNT(*) FILTER (WHERE e.status = 'queued') AS queue_length,
+                    COUNT(*) FILTER (WHERE e.status = 'active') AS active_count,
+                    MIN(e.enqueued_at) FILTER (WHERE e.status = 'queued') AS oldest_enqueued_at
+                FROM execution_admission_state s
+                LEFT JOIN execution_admission_entry e ON e.state_id = s.id
+                WHERE s.action_id = $1
+            )
+            SELECT
+                sr.state_count,
+                er.queue_length,
+                er.active_count,
+                sr.max_concurrent,
+                er.oldest_enqueued_at,
+                sr.total_enqueued,
+                sr.total_completed
+            FROM state_rows sr
+            CROSS JOIN entry_rows er
+            "#,
+        )
+        .bind(action_id)
+        .fetch_one(pool)
+        .await?;
+
+        let state_count: i64 = row.try_get("state_count")?;
+        if state_count == 0 {
+            return Ok(None);
+        }
+
+        Ok(Some(AdmissionQueueStats {
+            action_id,
+            queue_length: row.try_get::<i64, _>("queue_length")? as usize,
+            active_count: row.try_get::<i64, _>("active_count")? as u32,
+            max_concurrent: row.try_get::<i64, _>("max_concurrent")? as u32,
+            oldest_enqueued_at: row.try_get("oldest_enqueued_at")?,
+            total_enqueued: row.try_get::<i64, _>("total_enqueued")? as u64,
+            total_completed: row.try_get::<i64, _>("total_completed")? as u64,
+        }))
+    }
+
+    async fn enqueue_in_state(
+        tx: &mut Transaction<'_, Postgres>,
+        state: &AdmissionState,
+        max_queue_length: usize,
+        execution_id: Id,
+        allow_queue: bool,
+    ) -> Result<AdmissionEnqueueOutcome> {
+        if let Some(entry) = Self::find_execution_entry(tx, execution_id).await? {
+            if entry.status == "active" {
+                return Ok(AdmissionEnqueueOutcome::Acquired);
+            }
+
+            if entry.status == "queued" && entry.state_id == state.id {
+                if Self::maybe_promote_existing_queued(tx, state, execution_id).await? {
+                    return Ok(AdmissionEnqueueOutcome::Acquired);
+                }
+                return Ok(AdmissionEnqueueOutcome::Enqueued);
+            }
+
+            return Ok(AdmissionEnqueueOutcome::Enqueued);
+        }
+
+        let active_count = Self::active_count(tx, state.id).await?;
+        let queued_count = Self::queued_count(tx, state.id).await?;
+
+        if active_count < state.max_concurrent as i64 && queued_count == 0 {
+            let queue_order = Self::allocate_queue_order(tx, state.id).await?;
+            Self::insert_entry(
+                tx,
+                state.id,
+                execution_id,
+                "active",
+                queue_order,
+                Utc::now(),
+            )
+            .await?;
+            Self::increment_total_enqueued(tx, state.id).await?;
+            return Ok(AdmissionEnqueueOutcome::Acquired);
+        }
+
+        if !allow_queue {
+            return Ok(AdmissionEnqueueOutcome::Enqueued);
+        }
+
+        if queued_count >= max_queue_length as i64 {
+            return Err(anyhow::anyhow!(
+                "Queue full for action {}: maximum {} entries",
+                state.action_id,
+                max_queue_length
+            )
+            .into());
+        }
+
+        let queue_order = Self::allocate_queue_order(tx, state.id).await?;
+        Self::insert_entry(
+            tx,
+            state.id,
+            execution_id,
+            "queued",
+            queue_order,
+            Utc::now(),
+        )
+        .await?;
+        Self::increment_total_enqueued(tx, state.id).await?;
+        Ok(AdmissionEnqueueOutcome::Enqueued)
+    }
+
+    async fn maybe_promote_existing_queued(
+        tx: &mut Transaction<'_, Postgres>,
+        state: &AdmissionState,
+        execution_id: Id,
+    ) -> Result<bool> {
+        let active_count = Self::active_count(tx, state.id).await?;
+        if active_count >= state.max_concurrent as i64 {
+            return Ok(false);
+        }
+
+        let front_execution_id = sqlx::query_scalar::<Postgres, Id>(
+            r#"
+            SELECT execution_id
+            FROM execution_admission_entry
+            WHERE state_id = $1
+              AND status = 'queued'
+            ORDER BY queue_order ASC
+            LIMIT 1
+            "#,
+        )
+        .bind(state.id)
+        .fetch_optional(&mut **tx)
+        .await?;
+
+        if front_execution_id != Some(execution_id) {
+            return Ok(false);
+        }
+
+        sqlx::query(
+            r#"
+            UPDATE execution_admission_entry
+            SET status = 'active',
+                activated_at = NOW()
+            WHERE execution_id = $1
+              AND state_id = $2
+              AND status = 'queued'
+            "#,
+        )
+        .bind(execution_id)
+        .bind(state.id)
+        .execute(&mut **tx)
+        .await?;
+
+        Ok(true)
+    }
+
+    async fn promote_next_queued(
+        tx: &mut Transaction<'_, Postgres>,
+        state: &AdmissionState,
+    ) -> Result<Option<Id>> {
+        let next_execution_id = sqlx::query_scalar::<Postgres, Id>(
+            r#"
+            SELECT execution_id
+            FROM execution_admission_entry
+            WHERE state_id = $1
+              AND status = 'queued'
+            ORDER BY queue_order ASC
+            LIMIT 1
+            "#,
+        )
+        .bind(state.id)
+        .fetch_optional(&mut **tx)
+        .await?;
+
+        if let Some(next_execution_id) = next_execution_id {
+            sqlx::query(
+                r#"
+                UPDATE execution_admission_entry
+                SET status = 'active',
+                    activated_at = NOW()
+                WHERE execution_id = $1
+                  AND state_id = $2
+                  AND status = 'queued'
+                "#,
+            )
+            .bind(next_execution_id)
+            .bind(state.id)
+            .execute(&mut **tx)
+            .await?;
+        }
+
+        Ok(next_execution_id)
+    }
+
+    async fn lock_state(
+        tx: &mut Transaction<'_, Postgres>,
+        action_id: Id,
+        group_key: Option<String>,
+        max_concurrent: u32,
+    ) -> Result<AdmissionState> {
+        sqlx::query(
+            r#"
+            INSERT INTO execution_admission_state (action_id, group_key, max_concurrent)
+            VALUES ($1, $2, $3)
+            ON CONFLICT (action_id, group_key_normalized)
+            DO UPDATE SET max_concurrent = EXCLUDED.max_concurrent
+            "#,
+        )
+        .bind(action_id)
+        .bind(group_key.clone())
+        .bind(max_concurrent as i32)
+        .execute(&mut **tx)
+        .await?;
+
+        let state = sqlx::query(
+            r#"
+            SELECT id, action_id, group_key, max_concurrent
+            FROM execution_admission_state
+            WHERE action_id = $1
+              AND group_key_normalized = COALESCE($2, '')
+            FOR UPDATE
+            "#,
+        )
+        .bind(action_id)
+        .bind(group_key)
+        .fetch_one(&mut **tx)
+        .await?;
+
+        Ok(AdmissionState {
+            id: state.try_get("id")?,
+            action_id: state.try_get("action_id")?,
+            group_key: state.try_get("group_key")?,
+            max_concurrent: state.try_get("max_concurrent")?,
+        })
+    }
+
+    async fn lock_existing_state(
+        tx: &mut Transaction<'_, Postgres>,
+        action_id: Id,
+        group_key: Option<String>,
+    ) -> Result<Option<AdmissionState>> {
+        let row = sqlx::query(
+            r#"
+            SELECT id, action_id, group_key, max_concurrent
+            FROM execution_admission_state
+            WHERE action_id = $1
+              AND group_key_normalized = COALESCE($2, '')
+            FOR UPDATE
+            "#,
+        )
+        .bind(action_id)
+        .bind(group_key)
+        .fetch_optional(&mut **tx)
+        .await?;
+
+        Ok(row.map(|state| AdmissionState {
+            id: state.try_get("id").expect("state.id"),
+            action_id: state.try_get("action_id").expect("state.action_id"),
+            group_key: state.try_get("group_key").expect("state.group_key"),
+            max_concurrent: state
+                .try_get("max_concurrent")
+                .expect("state.max_concurrent"),
+        }))
+    }
+
+    async fn find_execution_entry(
+        tx: &mut Transaction<'_, Postgres>,
+        execution_id: Id,
+    ) -> Result<Option<ExecutionEntry>> {
+        let row = sqlx::query(
+            r#"
+            SELECT
+                e.state_id,
+                s.action_id,
+                s.group_key,
+                e.execution_id,
+                e.status,
+                e.queue_order,
+                e.enqueued_at
+            FROM execution_admission_entry e
+            JOIN execution_admission_state s ON s.id = e.state_id
+            WHERE e.execution_id = $1
+            "#,
+        )
+        .bind(execution_id)
+        .fetch_optional(&mut **tx)
+        .await?;
+
+        Ok(row.map(|entry| ExecutionEntry {
+            state_id: entry.try_get("state_id").expect("entry.state_id"),
+            action_id: entry.try_get("action_id").expect("entry.action_id"),
+            group_key: entry.try_get("group_key").expect("entry.group_key"),
+            status: entry.try_get("status").expect("entry.status"),
+            queue_order: entry.try_get("queue_order").expect("entry.queue_order"),
+            enqueued_at: entry.try_get("enqueued_at").expect("entry.enqueued_at"),
+        }))
+    }
+
+    async fn find_execution_entry_for_update(
+        tx: &mut Transaction<'_, Postgres>,
+        execution_id: Id,
+    ) -> Result<Option<ExecutionEntry>> {
+        let row = sqlx::query(
+            r#"
+            SELECT
+                e.state_id,
+                s.action_id,
+                s.group_key,
+                e.execution_id,
+                e.status,
+                e.queue_order,
+                e.enqueued_at
+            FROM execution_admission_entry e
+            JOIN execution_admission_state s ON s.id = e.state_id
+            WHERE e.execution_id = $1
+            FOR UPDATE OF e, s
+            "#,
+        )
+        .bind(execution_id)
+        .fetch_optional(&mut **tx)
+        .await?;
+
+        Ok(row.map(|entry| ExecutionEntry {
+            state_id: entry.try_get("state_id").expect("entry.state_id"),
+            action_id: entry.try_get("action_id").expect("entry.action_id"),
+            group_key: entry.try_get("group_key").expect("entry.group_key"),
+            status: entry.try_get("status").expect("entry.status"),
+            queue_order: entry.try_get("queue_order").expect("entry.queue_order"),
+            enqueued_at: entry.try_get("enqueued_at").expect("entry.enqueued_at"),
+        }))
+    }
+
+    async fn active_count(tx: &mut Transaction<'_, Postgres>, state_id: Id) -> Result<i64> {
+        Ok(sqlx::query_scalar::<Postgres, i64>(
+            r#"
+            SELECT COUNT(*)
+            FROM execution_admission_entry
+            WHERE state_id = $1
+              AND status = 'active'
+            "#,
+        )
+        .bind(state_id)
+        .fetch_one(&mut **tx)
+        .await?)
+    }
+
+    async fn queued_count(tx: &mut Transaction<'_, Postgres>, state_id: Id) -> Result<i64> {
+        Ok(sqlx::query_scalar::<Postgres, i64>(
+            r#"
+            SELECT COUNT(*)
+            FROM execution_admission_entry
+            WHERE state_id = $1
+              AND status = 'queued'
+            "#,
+        )
+        .bind(state_id)
+        .fetch_one(&mut **tx)
+        .await?)
+    }
+
+    async fn insert_entry(
+        tx: &mut Transaction<'_, Postgres>,
+        state_id: Id,
+        execution_id: Id,
+        status: &str,
+        queue_order: i64,
+        enqueued_at: DateTime<Utc>,
+    ) -> Result<()> {
+        sqlx::query(
+            r#"
+            INSERT INTO execution_admission_entry (
+                state_id, execution_id, status, queue_order, enqueued_at, activated_at
+            ) VALUES (
+                $1, $2, $3, $4, $5,
+                CASE WHEN $3 = 'active' THEN NOW() ELSE NULL END
+            )
+            "#,
+        )
+        .bind(state_id)
+        .bind(execution_id)
+        .bind(status)
+        .bind(queue_order)
+        .bind(enqueued_at)
+        .execute(&mut **tx)
+        .await?;
+
+        Ok(())
+    }
+
+    async fn allocate_queue_order(tx: &mut Transaction<'_, Postgres>, state_id: Id) -> Result<i64> {
+        let queue_order = sqlx::query_scalar::<Postgres, i64>(
+            r#"
+            UPDATE execution_admission_state
+            SET next_queue_order = next_queue_order + 1
+            WHERE id = $1
+            RETURNING next_queue_order - 1
+            "#,
+        )
+        .bind(state_id)
+        .fetch_one(&mut **tx)
+        .await?;
+
+        Ok(queue_order)
+    }
+
+    async fn increment_total_enqueued(
+        tx: &mut Transaction<'_, Postgres>,
+        state_id: Id,
+    ) -> Result<()> {
+        sqlx::query(
+            r#"
+            UPDATE execution_admission_state
+            SET total_enqueued = total_enqueued + 1
+            WHERE id = $1
+            "#,
+        )
+        .bind(state_id)
+        .execute(&mut **tx)
+        .await?;
+        Ok(())
+    }
+
+    async fn increment_total_completed(
+        tx: &mut Transaction<'_, Postgres>,
+        state_id: Id,
+    ) -> Result<()> {
+        sqlx::query(
+            r#"
+            UPDATE execution_admission_state
+            SET total_completed = total_completed + 1
+            WHERE id = $1
+            "#,
+        )
+        .bind(state_id)
+        .execute(&mut **tx)
+        .await?;
+        Ok(())
+    }
+
+    async fn refresh_queue_stats(tx: &mut Transaction<'_, Postgres>, action_id: Id) -> Result<()> {
+        let Some(stats) = Self::get_queue_stats_from_tx(tx, action_id).await? else {
+            QueueStatsRepository::delete(&mut **tx, action_id).await?;
+            return Ok(());
+        };
+
+        QueueStatsRepository::upsert(
+            &mut **tx,
+            UpsertQueueStatsInput {
+                action_id,
+                queue_length: stats.queue_length as i32,
+                active_count: stats.active_count as i32,
+                max_concurrent: stats.max_concurrent as i32,
+                oldest_enqueued_at: stats.oldest_enqueued_at,
+                total_enqueued: stats.total_enqueued as i64,
+                total_completed: stats.total_completed as i64,
+            },
+        )
+        .await?;
+
+        Ok(())
+    }
+
+    async fn get_queue_stats_from_tx(
+        tx: &mut Transaction<'_, Postgres>,
+        action_id: Id,
+    ) -> Result<Option<AdmissionQueueStats>> {
+        let row = sqlx::query(
+            r#"
+            WITH state_rows AS (
+                SELECT
+                    COUNT(*) AS state_count,
+                    COALESCE(SUM(max_concurrent), 0) AS max_concurrent,
+                    COALESCE(SUM(total_enqueued), 0) AS total_enqueued,
+                    COALESCE(SUM(total_completed), 0) AS total_completed
+                FROM execution_admission_state
+                WHERE action_id = $1
+            ),
+            entry_rows AS (
+                SELECT
+                    COUNT(*) FILTER (WHERE e.status = 'queued') AS queue_length,
+                    COUNT(*) FILTER (WHERE e.status = 'active') AS active_count,
+                    MIN(e.enqueued_at) FILTER (WHERE e.status = 'queued') AS oldest_enqueued_at
+                FROM execution_admission_state s
+                LEFT JOIN execution_admission_entry e ON e.state_id = s.id
+                WHERE s.action_id = $1
+            )
+            SELECT
+                sr.state_count,
+                er.queue_length,
+                er.active_count,
+                sr.max_concurrent,
+                er.oldest_enqueued_at,
+                sr.total_enqueued,
+                sr.total_completed
+            FROM state_rows sr
+            CROSS JOIN entry_rows er
+            "#,
+        )
+        .bind(action_id)
+        .fetch_one(&mut **tx)
+        .await?;
+
+        let state_count: i64 = row.try_get("state_count")?;
+        if state_count == 0 {
+            return Ok(None);
+        }
+
+        Ok(Some(AdmissionQueueStats {
+            action_id,
+            queue_length: row.try_get::<i64, _>("queue_length")? as usize,
+            active_count: row.try_get::<i64, _>("active_count")? as u32,
+            max_concurrent: row.try_get::<i64, _>("max_concurrent")? as u32,
+            oldest_enqueued_at: row.try_get("oldest_enqueued_at")?,
+            total_enqueued: row.try_get::<i64, _>("total_enqueued")? as u64,
+            total_completed: row.try_get::<i64, _>("total_completed")? as u64,
+        }))
+    }
+}

crates/common/src/repositories/mod.rs

@@ -33,6 +33,7 @@ pub mod artifact;
 pub mod entity_history;
 pub mod event;
 pub mod execution;
+pub mod execution_admission;
 pub mod identity;
 pub mod inquiry;
 pub mod key;
@@ -53,6 +54,7 @@ pub use artifact::{ArtifactRepository, ArtifactVersionRepository};
 pub use entity_history::EntityHistoryRepository;
 pub use event::{EnforcementRepository, EventRepository};
 pub use execution::ExecutionRepository;
+pub use execution_admission::ExecutionAdmissionRepository;
 pub use identity::{IdentityRepository, PermissionAssignmentRepository, PermissionSetRepository};
 pub use inquiry::InquiryRepository;
 pub use key::KeyRepository;

crates/common/src/repositories/queue_stats.rs

@@ -3,7 +3,7 @@
 //! Provides database operations for queue statistics persistence.
 
 use chrono::{DateTime, Utc};
-use sqlx::{PgPool, Postgres, QueryBuilder};
+use sqlx::{Executor, PgPool, Postgres, QueryBuilder};
 
 use crate::error::Result;
 use crate::models::Id;
@@ -38,7 +38,10 @@ pub struct QueueStatsRepository;
 
 impl QueueStatsRepository {
     /// Upsert queue statistics (insert or update)
-    pub async fn upsert(pool: &PgPool, input: UpsertQueueStatsInput) -> Result<QueueStats> {
+    pub async fn upsert<'e, E>(executor: E, input: UpsertQueueStatsInput) -> Result<QueueStats>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
         let stats = sqlx::query_as::<Postgres, QueueStats>(
             r#"
             INSERT INTO queue_stats (
@@ -69,14 +72,17 @@ impl QueueStatsRepository {
         .bind(input.oldest_enqueued_at)
         .bind(input.total_enqueued)
         .bind(input.total_completed)
-        .fetch_one(pool)
+        .fetch_one(executor)
         .await?;
 
         Ok(stats)
     }
 
     /// Get queue statistics for a specific action
-    pub async fn find_by_action(pool: &PgPool, action_id: Id) -> Result<Option<QueueStats>> {
+    pub async fn find_by_action<'e, E>(executor: E, action_id: Id) -> Result<Option<QueueStats>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
         let stats = sqlx::query_as::<Postgres, QueueStats>(
             r#"
             SELECT
@@ -93,14 +99,17 @@ impl QueueStatsRepository {
             "#,
         )
         .bind(action_id)
-        .fetch_optional(pool)
+        .fetch_optional(executor)
         .await?;
 
         Ok(stats)
     }
 
     /// List all queue statistics with active queues (queue_length > 0 or active_count > 0)
-    pub async fn list_active(pool: &PgPool) -> Result<Vec<QueueStats>> {
+    pub async fn list_active<'e, E>(executor: E) -> Result<Vec<QueueStats>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
         let stats = sqlx::query_as::<Postgres, QueueStats>(
             r#"
             SELECT
@@ -117,14 +126,17 @@ impl QueueStatsRepository {
             ORDER BY last_updated DESC
             "#,
         )
-        .fetch_all(pool)
+        .fetch_all(executor)
         .await?;
 
         Ok(stats)
     }
 
     /// List all queue statistics
-    pub async fn list_all(pool: &PgPool) -> Result<Vec<QueueStats>> {
+    pub async fn list_all<'e, E>(executor: E) -> Result<Vec<QueueStats>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
         let stats = sqlx::query_as::<Postgres, QueueStats>(
             r#"
             SELECT
@@ -140,14 +152,17 @@ impl QueueStatsRepository {
             ORDER BY last_updated DESC
             "#,
         )
-        .fetch_all(pool)
+        .fetch_all(executor)
         .await?;
 
         Ok(stats)
     }
 
     /// Delete queue statistics for a specific action
-    pub async fn delete(pool: &PgPool, action_id: Id) -> Result<bool> {
+    pub async fn delete<'e, E>(executor: E, action_id: Id) -> Result<bool>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
         let result = sqlx::query(
             r#"
             DELETE FROM queue_stats
@@ -155,7 +170,7 @@ impl QueueStatsRepository {
             "#,
         )
         .bind(action_id)
-        .execute(pool)
+        .execute(executor)
         .await?;
 
         Ok(result.rows_affected() > 0)
@@ -163,7 +178,7 @@ impl QueueStatsRepository {
 
     /// Batch upsert multiple queue statistics
     pub async fn batch_upsert(
-        pool: &PgPool,
+        executor: &PgPool,
         inputs: Vec<UpsertQueueStatsInput>,
     ) -> Result<Vec<QueueStats>> {
         if inputs.is_empty() {
@@ -213,14 +228,17 @@ impl QueueStatsRepository {
 
         let stats = query_builder
             .build_query_as::<QueueStats>()
-            .fetch_all(pool)
+            .fetch_all(executor)
             .await?;
 
         Ok(stats)
     }
 
     /// Clear stale statistics (older than specified duration)
-    pub async fn clear_stale(pool: &PgPool, older_than_seconds: i64) -> Result<u64> {
+    pub async fn clear_stale<'e, E>(executor: E, older_than_seconds: i64) -> Result<u64>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
         let result = sqlx::query(
             r#"
             DELETE FROM queue_stats
@@ -230,7 +248,7 @@ impl QueueStatsRepository {
             "#,
         )
         .bind(older_than_seconds)
-        .execute(pool)
+        .execute(executor)
         .await?;
 
         Ok(result.rows_affected())

crates/common/src/repositories/workflow.rs

@@ -411,6 +411,12 @@ impl WorkflowDefinitionRepository {
 
 pub struct WorkflowExecutionRepository;
 
+#[derive(Debug, Clone)]
+pub struct WorkflowExecutionCreateOrGetResult {
+    pub workflow_execution: WorkflowExecution,
+    pub created: bool,
+}
+
 impl Repository for WorkflowExecutionRepository {
     type Entity = WorkflowExecution;
     fn table_name() -> &'static str {
@@ -606,6 +612,71 @@ impl Delete for WorkflowExecutionRepository {
 }
 
 impl WorkflowExecutionRepository {
+    pub async fn find_by_id_for_update<'e, E>(
+        executor: E,
+        id: Id,
+    ) -> Result<Option<WorkflowExecution>>
+    where
+        E: Executor<'e, Database = Postgres> + 'e,
+    {
+        sqlx::query_as::<_, WorkflowExecution>(
+            "SELECT id, execution, workflow_def, current_tasks, completed_tasks, failed_tasks, skipped_tasks,
+                    variables, task_graph, status, error_message, paused, pause_reason, created, updated
+             FROM workflow_execution
+             WHERE id = $1
+             FOR UPDATE"
+        )
+        .bind(id)
+        .fetch_optional(executor)
+        .await
+        .map_err(Into::into)
+    }
+
+    pub async fn create_or_get_by_execution<'e, E>(
+        executor: E,
+        input: CreateWorkflowExecutionInput,
+    ) -> Result<WorkflowExecutionCreateOrGetResult>
+    where
+        E: Executor<'e, Database = Postgres> + Copy + 'e,
+    {
+        let inserted = sqlx::query_as::<_, WorkflowExecution>(
+            "INSERT INTO workflow_execution
+             (execution, workflow_def, task_graph, variables, status)
+             VALUES ($1, $2, $3, $4, $5)
+             ON CONFLICT (execution) DO NOTHING
+             RETURNING id, execution, workflow_def, current_tasks, completed_tasks, failed_tasks, skipped_tasks,
+                       variables, task_graph, status, error_message, paused, pause_reason, created, updated"
+        )
+        .bind(input.execution)
+        .bind(input.workflow_def)
+        .bind(&input.task_graph)
+        .bind(&input.variables)
+        .bind(input.status)
+        .fetch_optional(executor)
+        .await?;
+
+        if let Some(workflow_execution) = inserted {
+            return Ok(WorkflowExecutionCreateOrGetResult {
+                workflow_execution,
+                created: true,
+            });
+        }
+
+        let workflow_execution = Self::find_by_execution(executor, input.execution)
+            .await?
+            .ok_or_else(|| {
+                anyhow::anyhow!(
+                    "workflow_execution for parent execution {} disappeared after conflict",
+                    input.execution
+                )
+            })?;
+
+        Ok(WorkflowExecutionCreateOrGetResult {
+            workflow_execution,
+            created: false,
+        })
+    }
+
     /// Find workflow execution by the parent execution ID
     pub async fn find_by_execution<'e, E>(
         executor: E,

crates/common/tests/execution_repository_tests.rs

@@ -1153,3 +1153,108 @@ async fn test_execution_result_json() {
 
     assert_eq!(updated.result, Some(complex_result));
 }
+
+#[tokio::test]
+#[ignore = "integration test — requires database"]
+async fn test_claim_for_scheduling_succeeds_once() {
+    let pool = create_test_pool().await.unwrap();
+
+    let pack = PackFixture::new_unique("claim_pack")
+        .create(&pool)
+        .await
+        .unwrap();
+    let action = ActionFixture::new_unique(pack.id, &pack.r#ref, "claim_action")
+        .create(&pool)
+        .await
+        .unwrap();
+
+    let created = ExecutionRepository::create(
+        &pool,
+        CreateExecutionInput {
+            action: Some(action.id),
+            action_ref: action.r#ref.clone(),
+            config: None,
+            env_vars: None,
+            parent: None,
+            enforcement: None,
+            executor: None,
+            worker: None,
+            status: ExecutionStatus::Requested,
+            result: None,
+            workflow_task: None,
+        },
+    )
+    .await
+    .unwrap();
+
+    let first = ExecutionRepository::claim_for_scheduling(&pool, created.id, None)
+        .await
+        .unwrap();
+    let second = ExecutionRepository::claim_for_scheduling(&pool, created.id, None)
+        .await
+        .unwrap();
+
+    assert_eq!(first.unwrap().status, ExecutionStatus::Scheduling);
+    assert!(second.is_none());
+}
+
+#[tokio::test]
+#[ignore = "integration test — requires database"]
+async fn test_update_if_status_only_updates_matching_row() {
+    let pool = create_test_pool().await.unwrap();
+
+    let pack = PackFixture::new_unique("conditional_pack")
+        .create(&pool)
+        .await
+        .unwrap();
+    let action = ActionFixture::new_unique(pack.id, &pack.r#ref, "conditional_action")
+        .create(&pool)
+        .await
+        .unwrap();
+
+    let created = ExecutionRepository::create(
+        &pool,
+        CreateExecutionInput {
+            action: Some(action.id),
+            action_ref: action.r#ref.clone(),
+            config: None,
+            env_vars: None,
+            parent: None,
+            enforcement: None,
+            executor: None,
+            worker: None,
+            status: ExecutionStatus::Scheduling,
+            result: None,
+            workflow_task: None,
+        },
+    )
+    .await
+    .unwrap();
+
+    let updated = ExecutionRepository::update_if_status(
+        &pool,
+        created.id,
+        ExecutionStatus::Scheduling,
+        UpdateExecutionInput {
+            status: Some(ExecutionStatus::Scheduled),
+            worker: Some(77),
+            ..Default::default()
+        },
+    )
+    .await
+    .unwrap();
+    let skipped = ExecutionRepository::update_if_status(
+        &pool,
+        created.id,
+        ExecutionStatus::Scheduling,
+        UpdateExecutionInput {
+            status: Some(ExecutionStatus::Failed),
+            ..Default::default()
+        },
+    )
+    .await
+    .unwrap();
+
+    assert_eq!(updated.unwrap().status, ExecutionStatus::Scheduled);
+    assert!(skipped.is_none());
+}

crates/executor/src/completion_listener.rs

@@ -157,7 +157,11 @@ impl CompletionListener {
                         "Failed to advance workflow for execution {}: {}",
                         execution_id, e
                     );
-                    // Continue processing — don't fail the entire completion
+                    if let Some(mq_err) = Self::retryable_mq_error(&e) {
+                        return Err(mq_err.into());
+                    }
+                    // Non-retryable workflow advancement errors are logged but
+                    // do not fail the entire completion processing path.
                 }
             }
 

crates/executor/src/dead_letter_handler.rs

@@ -14,7 +14,7 @@ use attune_common::{
     error::Error,
     models::ExecutionStatus,
     mq::{Consumer, ConsumerConfig, MessageEnvelope, MessageType, MqResult},
-    repositories::{execution::UpdateExecutionInput, ExecutionRepository, FindById, Update},
+    repositories::{execution::UpdateExecutionInput, ExecutionRepository, FindById},
 };
 use chrono::Utc;
 use serde_json::json;
@@ -179,13 +179,12 @@ async fn handle_execution_requested(
         }
     };
 
-    // Only fail if still in a non-terminal state
-    if !matches!(
-        execution.status,
-        ExecutionStatus::Scheduled | ExecutionStatus::Running
-    ) {
+    // Only scheduled executions are still legitimately owned by the scheduler.
+    // If the execution already moved to running or a terminal state, this DLQ
+    // delivery is stale and must not overwrite newer state.
+    if execution.status != ExecutionStatus::Scheduled {
         info!(
-            "Execution {} already in terminal state {:?}, skipping",
+            "Execution {} already left Scheduled state ({:?}), skipping stale DLQ handling",
             execution_id, execution.status
         );
         return Ok(()); // Acknowledge to remove from queue
@@ -193,6 +192,12 @@ async fn handle_execution_requested(
 
     // Get worker info from payload for better error message
     let worker_id = envelope.payload.get("worker_id").and_then(|v| v.as_i64());
+    let scheduled_attempt_updated_at = envelope
+        .payload
+        .get("scheduled_attempt_updated_at")
+        .and_then(|v| v.as_str())
+        .and_then(|s| chrono::DateTime::parse_from_rfc3339(s).ok())
+        .map(|dt| dt.with_timezone(&Utc));
 
     let error_message = if let Some(wid) = worker_id {
         format!(
@@ -214,24 +219,85 @@ async fn handle_execution_requested(
         ..Default::default()
     };
 
-    match ExecutionRepository::update(pool, execution_id, update_input).await {
-        Ok(_) => {
-            info!(
-                "Successfully failed execution {} due to worker queue expiration",
-                execution_id
-            );
-            Ok(())
+    if let Some(timestamp) = scheduled_attempt_updated_at {
+        // Guard on both status and the exact updated_at from when the execution was
+        // scheduled — prevents overwriting state that changed after this DLQ message
+        // was enqueued.
+        match ExecutionRepository::update_if_status_and_updated_at(
+            pool,
+            execution_id,
+            ExecutionStatus::Scheduled,
+            timestamp,
+            update_input,
+        )
+        .await
+        {
+            Ok(Some(_)) => {
+                info!(
+                    "Successfully failed execution {} due to worker queue expiration",
+                    execution_id
+                );
+                Ok(())
+            }
+            Ok(None) => {
+                info!(
+                    "Skipping DLQ failure for execution {} because it already left Scheduled state",
+                    execution_id
+                );
+                Ok(())
+            }
+            Err(e) => {
+                error!(
+                    "Failed to update execution {} to failed state: {}",
+                    execution_id, e
+                );
+                Err(attune_common::mq::MqError::Consume(format!(
+                    "Failed to update execution: {}",
+                    e
+                )))
+            }
         }
-        Err(e) => {
-            error!(
-                "Failed to update execution {} to failed state: {}",
-                execution_id, e
-            );
-            // Return error to nack and potentially retry
-            Err(attune_common::mq::MqError::Consume(format!(
-                "Failed to update execution: {}",
-                e
-            )))
+    } else {
+        // Fallback for DLQ messages that predate the scheduled_attempt_updated_at
+        // field. Use a status-only guard — same safety guarantee as the original code
+        // (never overwrites terminal or running state).
+        warn!(
+            "DLQ message for execution {} lacks scheduled_attempt_updated_at; \
+             falling back to status-only guard",
+            execution_id
+        );
+        match ExecutionRepository::update_if_status(
+            pool,
+            execution_id,
+            ExecutionStatus::Scheduled,
+            update_input,
+        )
+        .await
+        {
+            Ok(Some(_)) => {
+                info!(
+                    "Successfully failed execution {} due to worker queue expiration (status-only guard)",
+                    execution_id
+                );
+                Ok(())
+            }
+            Ok(None) => {
+                info!(
+                    "Skipping DLQ failure for execution {} because it already left Scheduled state",
+                    execution_id
+                );
+                Ok(())
+            }
+            Err(e) => {
+                error!(
+                    "Failed to update execution {} to failed state: {}",
+                    execution_id, e
+                );
+                Err(attune_common::mq::MqError::Consume(format!(
+                    "Failed to update execution: {}",
+                    e
+                )))
+            }
         }
     }
 }

crates/executor/src/enforcement_processor.rs

@@ -19,7 +19,7 @@ use attune_common::{
         event::{EnforcementRepository, EventRepository, UpdateEnforcementInput},
         execution::{CreateExecutionInput, ExecutionRepository},
         rule::RuleRepository,
-        Create, FindById,
+        FindById,
     },
 };
 
@@ -116,6 +116,14 @@ impl EnforcementProcessor {
             .await?
             .ok_or_else(|| anyhow::anyhow!("Enforcement not found: {}", enforcement_id))?;
 
+        if enforcement.status != EnforcementStatus::Created {
+            debug!(
+                "Enforcement {} already left Created state ({:?}), skipping duplicate processing",
+                enforcement_id, enforcement.status
+            );
+            return Ok(());
+        }
+
         // Fetch associated rule
         let rule = RuleRepository::find_by_id(
             pool,
@@ -135,7 +143,7 @@ impl EnforcementProcessor {
 
         // Evaluate whether to create execution
         if Self::should_create_execution(&enforcement, &rule, event.as_ref())? {
-            Self::create_execution(
+            let execution_created = Self::create_execution(
                 pool,
                 publisher,
                 policy_enforcer,
@@ -145,10 +153,10 @@ impl EnforcementProcessor {
             )
             .await?;
 
-            // Update enforcement status to Processed after successful execution creation
-            EnforcementRepository::update_loaded(
+            let updated = EnforcementRepository::update_loaded_if_status(
                 pool,
                 &enforcement,
+                EnforcementStatus::Created,
                 UpdateEnforcementInput {
                     status: Some(EnforcementStatus::Processed),
                     payload: None,
@@ -157,17 +165,27 @@ impl EnforcementProcessor {
             )
             .await?;
 
-            debug!("Updated enforcement {} status to Processed", enforcement_id);
+            if updated.is_some() {
+                debug!(
+                    "Updated enforcement {} status to Processed after {} execution path",
+                    enforcement_id,
+                    if execution_created {
+                        "new"
+                    } else {
+                        "idempotent"
+                    }
+                );
+            }
         } else {
             info!(
                 "Skipping execution creation for enforcement: {}",
                 enforcement_id
             );
 
-            // Update enforcement status to Disabled since it was not actionable
-            EnforcementRepository::update_loaded(
+            let updated = EnforcementRepository::update_loaded_if_status(
                 pool,
                 &enforcement,
+                EnforcementStatus::Created,
                 UpdateEnforcementInput {
                     status: Some(EnforcementStatus::Disabled),
                     payload: None,
@@ -176,10 +194,12 @@ impl EnforcementProcessor {
             )
             .await?;
 
-            debug!(
-                "Updated enforcement {} status to Disabled (skipped)",
-                enforcement_id
-            );
+            if updated.is_some() {
+                debug!(
+                    "Updated enforcement {} status to Disabled (skipped)",
+                    enforcement_id
+                );
+            }
         }
 
         Ok(())
@@ -234,7 +254,7 @@ impl EnforcementProcessor {
         _queue_manager: &ExecutionQueueManager,
         enforcement: &Enforcement,
         rule: &Rule,
-    ) -> Result<()> {
+    ) -> Result<bool> {
         // Extract action ID — should_create_execution already verified it's Some,
         // but guard defensively here as well.
         let action_id = match rule.action {
@@ -275,44 +295,60 @@ impl EnforcementProcessor {
             workflow_task: None, // Non-workflow execution
         };
 
-        let execution = ExecutionRepository::create(pool, execution_input).await?;
+        let execution_result = ExecutionRepository::create_top_level_for_enforcement_if_absent(
+            pool,
+            execution_input,
+            enforcement.id,
+        )
+        .await?;
+        let execution = execution_result.execution;
 
-        info!(
-            "Created execution: {} for enforcement: {}",
-            execution.id, enforcement.id
-        );
+        if execution_result.created {
+            info!(
+                "Created execution: {} for enforcement: {}",
+                execution.id, enforcement.id
+            );
+        } else {
+            info!(
+                "Reusing execution: {} for enforcement: {}",
+                execution.id, enforcement.id
+            );
+        }
 
-        // Publish ExecutionRequested message
-        let payload = ExecutionRequestedPayload {
-            execution_id: execution.id,
-            action_id: Some(action_id),
-            action_ref: action_ref.clone(),
-            parent_id: None,
-            enforcement_id: Some(enforcement.id),
-            config: enforcement.config.clone(),
-        };
+        if execution_result.created
+            || execution.status == attune_common::models::enums::ExecutionStatus::Requested
+        {
+            let payload = ExecutionRequestedPayload {
+                execution_id: execution.id,
+                action_id: Some(action_id),
+                action_ref: action_ref.clone(),
+                parent_id: None,
+                enforcement_id: Some(enforcement.id),
+                config: execution.config.clone(),
+            };
 
-        let envelope =
-            MessageEnvelope::new(attune_common::mq::MessageType::ExecutionRequested, payload)
-                .with_source("executor");
+            let envelope =
+                MessageEnvelope::new(attune_common::mq::MessageType::ExecutionRequested, payload)
+                    .with_source("executor");
 
-        // Publish to execution requests queue with routing key
-        let routing_key = "execution.requested";
-        let exchange = "attune.executions";
+            // Publish to execution requests queue with routing key
+            let routing_key = "execution.requested";
+            let exchange = "attune.executions";
 
-        publisher
-            .publish_envelope_with_routing(&envelope, exchange, routing_key)
-            .await?;
+            publisher
+                .publish_envelope_with_routing(&envelope, exchange, routing_key)
+                .await?;
 
-        info!(
-            "Published execution.requested message for execution: {} (enforcement: {}, action: {})",
-            execution.id, enforcement.id, action_id
-        );
+            info!(
+                "Published execution.requested message for execution: {} (enforcement: {}, action: {})",
+                execution.id, enforcement.id, action_id
+            );
+        }
 
         // NOTE: Queue slot will be released when worker publishes execution.completed
         // and CompletionListener calls queue_manager.notify_completion(action_id)
 
-        Ok(())
+        Ok(execution_result.created)
     }
 }
 

crates/executor/src/event_processor.rs

@@ -19,7 +19,7 @@ use attune_common::{
         event::{CreateEnforcementInput, EnforcementRepository, EventRepository},
         pack::PackRepository,
         rule::RuleRepository,
-        Create, FindById, List,
+        FindById, List,
     },
     template_resolver::{resolve_templates, TemplateContext},
 };
@@ -206,32 +206,43 @@ impl EventProcessor {
             conditions: rule.conditions.clone(),
         };
 
-        let enforcement = EnforcementRepository::create(pool, create_input).await?;
+        let enforcement_result =
+            EnforcementRepository::create_or_get_by_rule_event(pool, create_input).await?;
+        let enforcement = enforcement_result.enforcement;
 
-        info!(
-            "Enforcement {} created for rule {} (event: {})",
-            enforcement.id, rule.r#ref, event.id
-        );
+        if enforcement_result.created {
+            info!(
+                "Enforcement {} created for rule {} (event: {})",
+                enforcement.id, rule.r#ref, event.id
+            );
+        } else {
+            info!(
+                "Reusing enforcement {} for rule {} (event: {})",
+                enforcement.id, rule.r#ref, event.id
+            );
+        }
 
-        // Publish EnforcementCreated message
-        let enforcement_payload = EnforcementCreatedPayload {
-            enforcement_id: enforcement.id,
-            rule_id: Some(rule.id),
-            rule_ref: rule.r#ref.clone(),
-            event_id: Some(event.id),
-            trigger_ref: event.trigger_ref.clone(),
-            payload: payload.clone(),
-        };
+        if enforcement_result.created || enforcement.status == EnforcementStatus::Created {
+            let enforcement_payload = EnforcementCreatedPayload {
+                enforcement_id: enforcement.id,
+                rule_id: Some(rule.id),
+                rule_ref: rule.r#ref.clone(),
+                event_id: Some(event.id),
+                trigger_ref: event.trigger_ref.clone(),
+                payload: payload.clone(),
+            };
 
-        let envelope = MessageEnvelope::new(MessageType::EnforcementCreated, enforcement_payload)
-            .with_source("event-processor");
+            let envelope =
+                MessageEnvelope::new(MessageType::EnforcementCreated, enforcement_payload)
+                    .with_source("event-processor");
 
-        publisher.publish_envelope(&envelope).await?;
+            publisher.publish_envelope(&envelope).await?;
 
-        debug!(
-            "Published EnforcementCreated message for enforcement {}",
-            enforcement.id
-        );
+            debug!(
+                "Published EnforcementCreated message for enforcement {}",
+                enforcement.id
+            );
+        }
 
         Ok(())
     }

crates/executor/src/inquiry_handler.rs

@@ -9,13 +9,14 @@
 
 use anyhow::Result;
 use attune_common::{
+    error::Error as AttuneError,
     models::{enums::InquiryStatus, inquiry::Inquiry, Execution, Id},
     mq::{
         Consumer, InquiryCreatedPayload, InquiryRespondedPayload, MessageEnvelope, MessageType,
         Publisher,
     },
     repositories::{
-        execution::{ExecutionRepository, UpdateExecutionInput},
+        execution::{ExecutionRepository, UpdateExecutionInput, SELECT_COLUMNS},
         inquiry::{CreateInquiryInput, InquiryRepository},
         Create, FindById, Update,
     },
@@ -28,6 +29,8 @@ use tracing::{debug, error, info, warn};
 
 /// Special key in action result to indicate an inquiry should be created
 pub const INQUIRY_RESULT_KEY: &str = "__inquiry";
+const INQUIRY_ID_RESULT_KEY: &str = "__inquiry_id";
+const INQUIRY_CREATED_PUBLISHED_RESULT_KEY: &str = "__inquiry_created_published";
 
 /// Structure for inquiry data in action results
 #[derive(Debug, Clone, serde::Deserialize)]
@@ -104,69 +107,198 @@ impl InquiryHandler {
         let inquiry_request: InquiryRequest = serde_json::from_value(inquiry_value.clone())?;
         Ok(inquiry_request)
     }
+}
 
+/// Returns true when `e` represents a PostgreSQL unique constraint violation (code 23505).
+fn is_db_unique_violation(e: &AttuneError) -> bool {
+    if let AttuneError::Database(sqlx_err) = e {
+        return sqlx_err
+            .as_database_error()
+            .and_then(|db| db.code())
+            .as_deref()
+            == Some("23505");
+    }
+    false
+}
+
+impl InquiryHandler {
     /// Create an inquiry for an execution and pause it
     pub async fn create_inquiry_from_result(
         pool: &PgPool,
         publisher: &Publisher,
         execution_id: Id,
-        result: &JsonValue,
+        _result: &JsonValue,
     ) -> Result<Inquiry> {
         info!("Creating inquiry for execution {}", execution_id);
 
-        // Extract inquiry request
-        let inquiry_request = Self::extract_inquiry_request(result)?;
+        let mut tx = pool.begin().await?;
+        let execution = sqlx::query_as::<_, Execution>(&format!(
+            "SELECT {SELECT_COLUMNS} FROM execution WHERE id = $1 FOR UPDATE"
+        ))
+        .bind(execution_id)
+        .fetch_one(&mut *tx)
+        .await?;
 
-        // Calculate timeout if specified
+        let mut result = execution
+            .result
+            .clone()
+            .ok_or_else(|| anyhow::anyhow!("Execution {} has no result", execution_id))?;
+        let inquiry_request = Self::extract_inquiry_request(&result)?;
         let timeout_at = inquiry_request
             .timeout_seconds
             .map(|seconds| Utc::now() + chrono::Duration::seconds(seconds));
 
-        // Create inquiry in database
-        let inquiry_input = CreateInquiryInput {
-            execution: execution_id,
-            prompt: inquiry_request.prompt.clone(),
-            response_schema: inquiry_request.response_schema.clone(),
-            assigned_to: inquiry_request.assigned_to,
-            status: InquiryStatus::Pending,
-            response: None,
-            timeout_at,
+        let existing_inquiry_id = result
+            .get(INQUIRY_ID_RESULT_KEY)
+            .and_then(|value| value.as_i64());
+        let published = result
+            .get(INQUIRY_CREATED_PUBLISHED_RESULT_KEY)
+            .and_then(|value| value.as_bool())
+            .unwrap_or(false);
+
+        let (inquiry, should_publish) = if let Some(inquiry_id) = existing_inquiry_id {
+            let inquiry = InquiryRepository::find_by_id(&mut *tx, inquiry_id)
+                .await?
+                .ok_or_else(|| {
+                    anyhow::anyhow!(
+                        "Inquiry {} referenced by execution {} result not found",
+                        inquiry_id,
+                        execution_id
+                    )
+                })?;
+            let should_publish = !published && inquiry.status == InquiryStatus::Pending;
+            (inquiry, should_publish)
+        } else {
+            let create_result = InquiryRepository::create(
+                &mut *tx,
+                CreateInquiryInput {
+                    execution: execution_id,
+                    prompt: inquiry_request.prompt.clone(),
+                    response_schema: inquiry_request.response_schema.clone(),
+                    assigned_to: inquiry_request.assigned_to,
+                    status: InquiryStatus::Pending,
+                    response: None,
+                    timeout_at,
+                },
+            )
+            .await;
+
+            let inquiry = match create_result {
+                Ok(inq) => inq,
+                Err(e) => {
+                    // Unique constraint violation (23505): another replica already
+                    // created the inquiry for this execution. Treat as idempotent
+                    // success — drop the aborted transaction and return the existing row.
+                    if is_db_unique_violation(&e) {
+                        info!(
+                            "Inquiry for execution {} already created by another replica \
+                             (unique constraint 23505); treating as idempotent",
+                            execution_id
+                        );
+                        // tx is in an aborted state; dropping it issues ROLLBACK.
+                        drop(tx);
+                        let inquiries =
+                            InquiryRepository::find_by_execution(pool, execution_id).await?;
+                        let existing = inquiries.into_iter().next().ok_or_else(|| {
+                            anyhow::anyhow!(
+                                "Inquiry for execution {} not found after unique constraint violation",
+                                execution_id
+                            )
+                        })?;
+                        return Ok(existing);
+                    }
+                    return Err(e.into());
+                }
+            };
+
+            Self::set_inquiry_result_metadata(&mut result, inquiry.id, false)?;
+            ExecutionRepository::update(
+                &mut *tx,
+                execution_id,
+                UpdateExecutionInput {
+                    result: Some(result),
+                    ..Default::default()
+                },
+            )
+            .await?;
+
+            (inquiry, true)
         };
 
-        let inquiry = InquiryRepository::create(pool, inquiry_input).await?;
+        tx.commit().await?;
 
-        info!(
-            "Created inquiry {} for execution {}",
-            inquiry.id, execution_id
-        );
+        if should_publish {
+            let payload = InquiryCreatedPayload {
+                inquiry_id: inquiry.id,
+                execution_id,
+                prompt: inquiry_request.prompt,
+                response_schema: inquiry_request.response_schema,
+                assigned_to: inquiry_request.assigned_to,
+                timeout_at,
+            };
 
-        // Update execution status to paused/waiting
-        // Note: We use a special status or keep it as "running" with inquiry tracking
-        // For now, we'll keep status as-is and track via inquiry relationship
+            let envelope =
+                MessageEnvelope::new(MessageType::InquiryCreated, payload).with_source("executor");
 
-        // Publish InquiryCreated message
-        let payload = InquiryCreatedPayload {
-            inquiry_id: inquiry.id,
-            execution_id,
-            prompt: inquiry_request.prompt,
-            response_schema: inquiry_request.response_schema,
-            assigned_to: inquiry_request.assigned_to,
-            timeout_at,
-        };
+            publisher.publish_envelope(&envelope).await?;
+            Self::mark_inquiry_created_published(pool, execution_id).await?;
 
-        let envelope =
-            MessageEnvelope::new(MessageType::InquiryCreated, payload).with_source("executor");
-
-        publisher.publish_envelope(&envelope).await?;
-
-        debug!(
-            "Published InquiryCreated message for inquiry {}",
-            inquiry.id
-        );
+            debug!(
+                "Published InquiryCreated message for inquiry {}",
+                inquiry.id
+            );
+        }
 
         Ok(inquiry)
     }
 
+    fn set_inquiry_result_metadata(
+        result: &mut JsonValue,
+        inquiry_id: Id,
+        published: bool,
+    ) -> Result<()> {
+        let obj = result
+            .as_object_mut()
+            .ok_or_else(|| anyhow::anyhow!("execution result is not a JSON object"))?;
+
+        obj.insert(
+            INQUIRY_ID_RESULT_KEY.to_string(),
+            JsonValue::Number(inquiry_id.into()),
+        );
+        obj.insert(
+            INQUIRY_CREATED_PUBLISHED_RESULT_KEY.to_string(),
+            JsonValue::Bool(published),
+        );
+        Ok(())
+    }
+
+    async fn mark_inquiry_created_published(pool: &PgPool, execution_id: Id) -> Result<()> {
+        let execution = ExecutionRepository::find_by_id(pool, execution_id)
+            .await?
+            .ok_or_else(|| anyhow::anyhow!("Execution {} not found", execution_id))?;
+        let mut result = execution
+            .result
+            .clone()
+            .ok_or_else(|| anyhow::anyhow!("Execution {} has no result", execution_id))?;
+        let inquiry_id = result
+            .get(INQUIRY_ID_RESULT_KEY)
+            .and_then(|value| value.as_i64())
+            .ok_or_else(|| anyhow::anyhow!("Execution {} missing __inquiry_id", execution_id))?;
+
+        Self::set_inquiry_result_metadata(&mut result, inquiry_id, true)?;
+        ExecutionRepository::update(
+            pool,
+            execution_id,
+            UpdateExecutionInput {
+                result: Some(result),
+                ..Default::default()
+            },
+        )
+        .await?;
+
+        Ok(())
+    }
+
     /// Handle an inquiry response message
     async fn handle_inquiry_response(
         pool: &PgPool,
@@ -235,9 +367,13 @@ impl InquiryHandler {
         if let Some(obj) = updated_result.as_object_mut() {
             obj.insert("__inquiry_response".to_string(), response.clone());
             obj.insert(
-                "__inquiry_id".to_string(),
+                INQUIRY_ID_RESULT_KEY.to_string(),
                 JsonValue::Number(inquiry.id.into()),
             );
+            obj.insert(
+                INQUIRY_CREATED_PUBLISHED_RESULT_KEY.to_string(),
+                JsonValue::Bool(true),
+            );
         }
 
         // Update execution with new result

crates/executor/src/policy_enforcer.rs

@@ -933,8 +933,8 @@ mod tests {
         assert_eq!(enforcer.get_concurrency_limit(2, Some(200)), Some(20));
     }
 
-    #[test]
-    fn test_build_parameter_group_key_uses_exact_values() {
+    #[tokio::test]
+    async fn test_build_parameter_group_key_uses_exact_values() {
         let pool = sqlx::PgPool::connect_lazy("postgresql://localhost/test").unwrap();
         let enforcer = PolicyEnforcer::new(pool);
         let config = serde_json::json!({

crates/executor/src/queue_manager.rs

@@ -23,7 +23,13 @@ use tokio::time::{sleep, Duration, Instant};
 use tracing::{debug, info, warn};
 
 use attune_common::models::Id;
-use attune_common::repositories::queue_stats::{QueueStatsRepository, UpsertQueueStatsInput};
+use attune_common::repositories::{
+    execution_admission::{
+        AdmissionEnqueueOutcome, AdmissionQueueStats, AdmissionQueuedRemovalOutcome,
+        AdmissionSlotReleaseOutcome, ExecutionAdmissionRepository,
+    },
+    queue_stats::{QueueStatsRepository, UpsertQueueStatsInput},
+};
 
 /// Configuration for the queue manager
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -51,6 +57,8 @@ impl Default for QueueConfig {
 struct QueueEntry {
     /// Execution or enforcement ID being queued
     execution_id: Id,
+    /// Durable FIFO position for the DB-backed admission path.
+    queue_order: Option<i64>,
     /// When this entry was added to the queue
     enqueued_at: DateTime<Utc>,
 }
@@ -224,6 +232,12 @@ impl ExecutionQueueManager {
         max_concurrent: u32,
         group_key: Option<String>,
     ) -> Result<()> {
+        if self.db_pool.is_some() {
+            return self
+                .enqueue_and_wait_db(action_id, execution_id, max_concurrent, group_key)
+                .await;
+        }
+
         if self.active_execution_keys.contains_key(&execution_id) {
             debug!(
                 "Execution {} already owns an active slot, skipping queue wait",
@@ -311,6 +325,7 @@ impl ExecutionQueueManager {
             // Add to queue
             let entry = QueueEntry {
                 execution_id,
+                queue_order: None,
                 enqueued_at: Utc::now(),
             };
 
@@ -392,6 +407,24 @@ impl ExecutionQueueManager {
         max_concurrent: u32,
         group_key: Option<String>,
     ) -> Result<SlotEnqueueOutcome> {
+        if let Some(pool) = &self.db_pool {
+            return Ok(
+                match ExecutionAdmissionRepository::enqueue(
+                    pool,
+                    self.config.max_queue_length,
+                    action_id,
+                    execution_id,
+                    max_concurrent,
+                    group_key,
+                )
+                .await?
+                {
+                    AdmissionEnqueueOutcome::Acquired => SlotEnqueueOutcome::Acquired,
+                    AdmissionEnqueueOutcome::Enqueued => SlotEnqueueOutcome::Enqueued,
+                },
+            );
+        }
+
         if self.active_execution_keys.contains_key(&execution_id) {
             debug!(
                 "Execution {} already owns an active slot, treating as acquired",
@@ -463,6 +496,7 @@ impl ExecutionQueueManager {
 
             queue.queue.push_back(QueueEntry {
                 execution_id,
+                queue_order: None,
                 enqueued_at: Utc::now(),
             });
             queue.total_enqueued += 1;
@@ -480,6 +514,21 @@ impl ExecutionQueueManager {
         max_concurrent: u32,
         group_key: Option<String>,
     ) -> Result<SlotAcquireOutcome> {
+        if let Some(pool) = &self.db_pool {
+            let outcome = ExecutionAdmissionRepository::try_acquire(
+                pool,
+                action_id,
+                execution_id,
+                max_concurrent,
+                group_key,
+            )
+            .await?;
+            return Ok(SlotAcquireOutcome {
+                acquired: outcome.acquired,
+                current_count: outcome.current_count,
+            });
+        }
+
         let queue_key = self.queue_key(action_id, group_key);
         let queue_arc = self
             .get_or_create_queue(queue_key.clone(), max_concurrent)
@@ -530,6 +579,14 @@ impl ExecutionQueueManager {
         &self,
         execution_id: Id,
     ) -> Result<Option<SlotReleaseOutcome>> {
+        if let Some(pool) = &self.db_pool {
+            return Ok(
+                ExecutionAdmissionRepository::release_active_slot(pool, execution_id)
+                    .await?
+                    .map(Self::map_release_outcome),
+            );
+        }
+
         let Some((_, queue_key)) = self.active_execution_keys.remove(&execution_id) else {
             debug!(
                 "No active queue slot found for execution {} (queue may have been cleared)",
@@ -610,6 +667,16 @@ impl ExecutionQueueManager {
         execution_id: Id,
         outcome: &SlotReleaseOutcome,
     ) -> Result<()> {
+        if let Some(pool) = &self.db_pool {
+            ExecutionAdmissionRepository::restore_active_slot(
+                pool,
+                execution_id,
+                &Self::to_admission_release_outcome(outcome),
+            )
+            .await?;
+            return Ok(());
+        }
+
         let action_id = outcome.queue_key.action_id;
         let queue_arc = self.get_or_create_queue(outcome.queue_key.clone(), 1).await;
         let mut queue = queue_arc.lock().await;
@@ -630,6 +697,14 @@ impl ExecutionQueueManager {
         &self,
         execution_id: Id,
     ) -> Result<Option<QueuedRemovalOutcome>> {
+        if let Some(pool) = &self.db_pool {
+            return Ok(
+                ExecutionAdmissionRepository::remove_queued_execution(pool, execution_id)
+                    .await?
+                    .map(Self::map_removal_outcome),
+            );
+        }
+
         for entry in self.queues.iter() {
             let queue_key = entry.key().clone();
             let queue_arc = entry.value().clone();
@@ -666,6 +741,15 @@ impl ExecutionQueueManager {
     }
 
     pub async fn restore_queued_execution(&self, outcome: &QueuedRemovalOutcome) -> Result<()> {
+        if let Some(pool) = &self.db_pool {
+            ExecutionAdmissionRepository::restore_queued_execution(
+                pool,
+                &Self::to_admission_removal_outcome(outcome),
+            )
+            .await?;
+            return Ok(());
+        }
+
         let action_id = outcome.queue_key.action_id;
         let queue_arc = self.get_or_create_queue(outcome.queue_key.clone(), 1).await;
         let mut queue = queue_arc.lock().await;
@@ -709,6 +793,19 @@ impl ExecutionQueueManager {
 
     /// Get statistics for a specific action's queue
     pub async fn get_queue_stats(&self, action_id: Id) -> Option<QueueStats> {
+        if let Some(pool) = &self.db_pool {
+            return ExecutionAdmissionRepository::get_queue_stats(pool, action_id)
+                .await
+                .map(|stats| stats.map(Self::map_queue_stats))
+                .unwrap_or_else(|err| {
+                    warn!(
+                        "Failed to load shared queue stats for action {}: {}",
+                        action_id, err
+                    );
+                    None
+                });
+        }
+
         let queue_arcs: Vec<Arc<Mutex<ActionQueue>>> = self
             .queues
             .iter()
@@ -757,6 +854,26 @@ impl ExecutionQueueManager {
     /// Get statistics for all queues
     #[allow(dead_code)]
     pub async fn get_all_queue_stats(&self) -> Vec<QueueStats> {
+        if let Some(pool) = &self.db_pool {
+            return QueueStatsRepository::list_all(pool)
+                .await
+                .map(|stats| {
+                    stats
+                        .into_iter()
+                        .map(|stat| QueueStats {
+                            action_id: stat.action_id,
+                            queue_length: stat.queue_length as usize,
+                            active_count: stat.active_count as u32,
+                            max_concurrent: stat.max_concurrent as u32,
+                            oldest_enqueued_at: stat.oldest_enqueued_at,
+                            total_enqueued: stat.total_enqueued as u64,
+                            total_completed: stat.total_completed as u64,
+                        })
+                        .collect()
+                })
+                .unwrap_or_default();
+        }
+
         let mut stats = Vec::new();
 
         let mut action_ids = std::collections::BTreeSet::new();
@@ -787,6 +904,14 @@ impl ExecutionQueueManager {
     /// * `Ok(false)` - Execution not found in queue
     #[allow(dead_code)]
     pub async fn cancel_execution(&self, action_id: Id, execution_id: Id) -> Result<bool> {
+        if let Some(pool) = &self.db_pool {
+            return Ok(
+                ExecutionAdmissionRepository::remove_queued_execution(pool, execution_id)
+                    .await?
+                    .is_some(),
+            );
+        }
+
         debug!(
             "Attempting to cancel execution {} for action {}",
             execution_id, action_id
@@ -838,12 +963,147 @@ impl ExecutionQueueManager {
     /// Get the number of actions with active queues
     #[allow(dead_code)]
     pub fn active_queue_count(&self) -> usize {
+        if self.db_pool.is_some() {
+            return 0;
+        }
+
         self.queues
             .iter()
             .map(|entry| entry.key().action_id)
             .collect::<std::collections::BTreeSet<_>>()
             .len()
     }
+
+    async fn enqueue_and_wait_db(
+        &self,
+        action_id: Id,
+        execution_id: Id,
+        max_concurrent: u32,
+        group_key: Option<String>,
+    ) -> Result<()> {
+        let pool = self
+            .db_pool
+            .as_ref()
+            .ok_or_else(|| anyhow::anyhow!("database pool required for shared admission"))?;
+
+        match ExecutionAdmissionRepository::enqueue(
+            pool,
+            self.config.max_queue_length,
+            action_id,
+            execution_id,
+            max_concurrent,
+            group_key.clone(),
+        )
+        .await?
+        {
+            AdmissionEnqueueOutcome::Acquired => return Ok(()),
+            AdmissionEnqueueOutcome::Enqueued => {}
+        }
+
+        let deadline = Instant::now() + Duration::from_secs(self.config.queue_timeout_seconds);
+        loop {
+            sleep(Duration::from_millis(10)).await;
+
+            match ExecutionAdmissionRepository::wait_status(pool, execution_id).await? {
+                Some(true) => return Ok(()),
+                Some(false) => {}
+                None => {
+                    return Err(anyhow::anyhow!(
+                        "Queue state for execution {} disappeared while waiting",
+                        execution_id
+                    ));
+                }
+            }
+
+            if Instant::now() < deadline {
+                continue;
+            }
+
+            match ExecutionAdmissionRepository::remove_queued_execution(pool, execution_id).await? {
+                Some(_) => {
+                    return Err(anyhow::anyhow!(
+                        "Queue timeout for execution {}: waited {} seconds",
+                        execution_id,
+                        self.config.queue_timeout_seconds
+                    ));
+                }
+                None => {
+                    if matches!(
+                        ExecutionAdmissionRepository::wait_status(pool, execution_id).await?,
+                        Some(true)
+                    ) {
+                        return Ok(());
+                    }
+
+                    return Err(anyhow::anyhow!(
+                        "Queue timeout for execution {}: waited {} seconds",
+                        execution_id,
+                        self.config.queue_timeout_seconds
+                    ));
+                }
+            }
+        }
+    }
+
+    fn map_release_outcome(outcome: AdmissionSlotReleaseOutcome) -> SlotReleaseOutcome {
+        SlotReleaseOutcome {
+            next_execution_id: outcome.next_execution_id,
+            queue_key: QueueKey {
+                action_id: outcome.action_id,
+                group_key: outcome.group_key,
+            },
+        }
+    }
+
+    fn to_admission_release_outcome(outcome: &SlotReleaseOutcome) -> AdmissionSlotReleaseOutcome {
+        AdmissionSlotReleaseOutcome {
+            action_id: outcome.queue_key.action_id,
+            group_key: outcome.queue_key.group_key.clone(),
+            next_execution_id: outcome.next_execution_id,
+        }
+    }
+
+    fn map_removal_outcome(outcome: AdmissionQueuedRemovalOutcome) -> QueuedRemovalOutcome {
+        QueuedRemovalOutcome {
+            next_execution_id: outcome.next_execution_id,
+            queue_key: QueueKey {
+                action_id: outcome.action_id,
+                group_key: outcome.group_key,
+            },
+            removed_entry: QueueEntry {
+                execution_id: outcome.execution_id,
+                queue_order: Some(outcome.queue_order),
+                enqueued_at: outcome.enqueued_at,
+            },
+            removed_index: outcome.removed_index,
+        }
+    }
+
+    fn to_admission_removal_outcome(
+        outcome: &QueuedRemovalOutcome,
+    ) -> AdmissionQueuedRemovalOutcome {
+        AdmissionQueuedRemovalOutcome {
+            action_id: outcome.queue_key.action_id,
+            group_key: outcome.queue_key.group_key.clone(),
+            next_execution_id: outcome.next_execution_id,
+            execution_id: outcome.removed_entry.execution_id,
+            queue_order: outcome.removed_entry.queue_order.unwrap_or_default(),
+            enqueued_at: outcome.removed_entry.enqueued_at,
+            removed_index: outcome.removed_index,
+        }
+    }
+
+    fn map_queue_stats(stats: AdmissionQueueStats) -> QueueStats {
+        QueueStats {
+            action_id: stats.action_id,
+            queue_length: stats.queue_length,
+            active_count: stats.active_count,
+            max_concurrent: stats.max_concurrent,
+            oldest_enqueued_at: stats.oldest_enqueued_at,
+            total_enqueued: stats.total_enqueued,
+            total_completed: stats.total_completed,
+        }
+    }
 }
 
 #[cfg(test)]

crates/executor/src/timeout_monitor.rs

@@ -12,7 +12,10 @@ use anyhow::Result;
 use attune_common::{
     models::{enums::ExecutionStatus, Execution},
     mq::{MessageEnvelope, MessageType, Publisher},
-    repositories::execution::SELECT_COLUMNS as EXECUTION_COLUMNS,
+    repositories::{
+        execution::{UpdateExecutionInput, SELECT_COLUMNS as EXECUTION_COLUMNS},
+        ExecutionRepository,
+    },
 };
 use chrono::{DateTime, Utc};
 use serde::{Deserialize, Serialize};
@@ -178,20 +181,27 @@ impl ExecutionTimeoutMonitor {
             "original_status": "scheduled"
         });
 
-        // Update execution status in database
-        sqlx::query(
-            "UPDATE execution
-             SET status = $1,
-                 result = $2,
-                 updated = NOW()
-             WHERE id = $3",
+        let updated = ExecutionRepository::update_if_status_and_updated_before(
+            &self.pool,
+            execution_id,
+            ExecutionStatus::Scheduled,
+            self.calculate_cutoff_time(),
+            UpdateExecutionInput {
+                status: Some(ExecutionStatus::Failed),
+                result: Some(result.clone()),
+                ..Default::default()
+            },
         )
-        .bind(ExecutionStatus::Failed)
-        .bind(&result)
-        .bind(execution_id)
-        .execute(&self.pool)
         .await?;
 
+        if updated.is_none() {
+            debug!(
+                "Skipping timeout failure for execution {} because it already left Scheduled or is no longer stale",
+                execution_id
+            );
+            return Ok(());
+        }
+
         info!("Execution {} marked as failed in database", execution_id);
 
         // Publish completion notification

crates/executor/tests/fifo_ordering_integration_test.rs

@@ -912,6 +912,115 @@ async fn test_queue_stats_persistence() {
     cleanup_test_data(&pool, pack_id).await;
 }
 
+#[tokio::test]
+#[ignore] // Requires database
+async fn test_release_restore_recovers_active_slot_and_next_queue_head() {
+    let pool = setup_db().await;
+    let timestamp = Utc::now().timestamp();
+    let suffix = format!("restore_release_{}", timestamp);
+
+    let pack_id = create_test_pack(&pool, &suffix).await;
+    let pack_ref = format!("fifo_test_pack_{}", suffix);
+    let action_id = create_test_action(&pool, pack_id, &pack_ref, &suffix).await;
+    let action_ref = format!("fifo_test_action_{}", suffix);
+
+    let manager = ExecutionQueueManager::with_db_pool(QueueConfig::default(), pool.clone());
+
+    let first =
+        create_test_execution(&pool, action_id, &action_ref, ExecutionStatus::Requested).await;
+    let second =
+        create_test_execution(&pool, action_id, &action_ref, ExecutionStatus::Requested).await;
+    let third =
+        create_test_execution(&pool, action_id, &action_ref, ExecutionStatus::Requested).await;
+
+    manager.enqueue(action_id, first, 1, None).await.unwrap();
+    manager.enqueue(action_id, second, 1, None).await.unwrap();
+    manager.enqueue(action_id, third, 1, None).await.unwrap();
+
+    let stats = manager.get_queue_stats(action_id).await.unwrap();
+    assert_eq!(stats.active_count, 1);
+    assert_eq!(stats.queue_length, 2);
+
+    let release = manager
+        .release_active_slot(first)
+        .await
+        .unwrap()
+        .expect("first execution should own an active slot");
+    assert_eq!(release.next_execution_id, Some(second));
+
+    let stats = manager.get_queue_stats(action_id).await.unwrap();
+    assert_eq!(stats.active_count, 1);
+    assert_eq!(stats.queue_length, 1);
+
+    manager.restore_active_slot(first, &release).await.unwrap();
+
+    let stats = manager.get_queue_stats(action_id).await.unwrap();
+    assert_eq!(stats.active_count, 1);
+    assert_eq!(stats.queue_length, 2);
+    assert_eq!(stats.total_completed, 0);
+
+    let next = manager
+        .release_active_slot(first)
+        .await
+        .unwrap()
+        .expect("restored execution should still own the active slot");
+    assert_eq!(next.next_execution_id, Some(second));
+
+    cleanup_test_data(&pool, pack_id).await;
+}
+
+#[tokio::test]
+#[ignore] // Requires database
+async fn test_remove_restore_recovers_queued_execution_position() {
+    let pool = setup_db().await;
+    let timestamp = Utc::now().timestamp();
+    let suffix = format!("restore_queue_{}", timestamp);
+
+    let pack_id = create_test_pack(&pool, &suffix).await;
+    let pack_ref = format!("fifo_test_pack_{}", suffix);
+    let action_id = create_test_action(&pool, pack_id, &pack_ref, &suffix).await;
+    let action_ref = format!("fifo_test_action_{}", suffix);
+
+    let manager = ExecutionQueueManager::with_db_pool(QueueConfig::default(), pool.clone());
+
+    let first =
+        create_test_execution(&pool, action_id, &action_ref, ExecutionStatus::Requested).await;
+    let second =
+        create_test_execution(&pool, action_id, &action_ref, ExecutionStatus::Requested).await;
+    let third =
+        create_test_execution(&pool, action_id, &action_ref, ExecutionStatus::Requested).await;
+
+    manager.enqueue(action_id, first, 1, None).await.unwrap();
+    manager.enqueue(action_id, second, 1, None).await.unwrap();
+    manager.enqueue(action_id, third, 1, None).await.unwrap();
+
+    let removal = manager
+        .remove_queued_execution(second)
+        .await
+        .unwrap()
+        .expect("second execution should be queued");
+    assert_eq!(removal.next_execution_id, None);
+
+    let stats = manager.get_queue_stats(action_id).await.unwrap();
+    assert_eq!(stats.active_count, 1);
+    assert_eq!(stats.queue_length, 1);
+
+    manager.restore_queued_execution(&removal).await.unwrap();
+
+    let stats = manager.get_queue_stats(action_id).await.unwrap();
+    assert_eq!(stats.active_count, 1);
+    assert_eq!(stats.queue_length, 2);
+
+    let release = manager
+        .release_active_slot(first)
+        .await
+        .unwrap()
+        .expect("first execution should own the active slot");
+    assert_eq!(release.next_execution_id, Some(second));
+
+    cleanup_test_data(&pool, pack_id).await;
+}
+
 #[tokio::test]
 #[ignore] // Requires database
 async fn test_queue_full_rejection() {

docs/plans/executor-ha-horizontal-scaling.md

@@ -0,0 +1,400 @@
+# Executor HA Horizontal Scaling Plan
+
+## Overview
+
+This plan describes the changes required to make the Attune executor service safe to run with multiple replicas. The current implementation already uses RabbitMQ competing consumers for most work distribution, but several correctness-critical parts of the executor still rely on process-local memory or non-atomic database updates. Those assumptions make the service unsafe under horizontal scaling, message replay, or partial failure.
+
+The goal of this plan is to make executor behavior correct under:
+
+- Multiple executor replicas running concurrently
+- Message redelivery from RabbitMQ
+- Replica crash/restart during scheduling or workflow advancement
+- Background recovery loops running on more than one replica
+
+## Problem Summary
+
+The current executor has five HA blockers:
+
+1. **Concurrency/FIFO control is process-local.**
+   `ExecutionQueueManager` stores active slots and waiting queues in memory. That means one replica can admit work while another replica receives the completion notification, causing slot release and queue advancement to fail.
+
+2. **Execution scheduling has no atomic claim step.**
+   The scheduler reads an execution in `requested`, does policy checks and worker selection, then updates it later. Two replicas can both observe the same row as schedulable and both dispatch it.
+
+3. **Workflow orchestration is not serialized.**
+   Workflow start and workflow advancement perform read-check-create-update sequences without a distributed lock or optimistic version check, so duplicate successor tasks can be created.
+
+4. **Event/enforcement/inquiry handlers are not idempotent.**
+   Duplicate message delivery can create duplicate enforcements, duplicate executions, duplicate workflow starts, and duplicate inquiries.
+
+5. **Timeout and DLQ handlers use non-conditional updates.**
+   Recovery loops can overwrite newer worker-owned state if the execution changes between the initial read and the final update.
+
+## Goals
+
+- Make execution scheduling single-winner under multiple executor replicas
+- Make policy-based concurrency control and FIFO queueing shared across replicas
+- Make workflow start and workflow advancement idempotent and serialized
+- Make duplicate message delivery safe for executor-owned entities
+- Make recovery loops safe to run on every replica
+
+## Non-Goals
+
+- Re-architecting RabbitMQ usage across the whole platform
+- Replacing PostgreSQL with a dedicated distributed lock service
+- Solving general worker autoscaling or Kubernetes deployment concerns in this phase
+- Reworking unrelated executor features like retry policy design unless needed for HA correctness
+
+## Design Principles
+
+### Database is the source of truth
+
+All coordination state that affects correctness must live in PostgreSQL, not in executor memory. In-memory caches and metrics are fine as optimization or observability layers, but correctness cannot depend on them.
+
+### State transitions must be compare-and-swap
+
+Any executor action that changes ownership or lifecycle state must use an atomic update that verifies the prior state in the same statement or transaction.
+
+### Handlers must be idempotent on domain keys
+
+RabbitMQ gives at-least-once delivery semantics. The executor must therefore tolerate duplicate delivery even when the same message is processed more than once or by different replicas.
+
+### Workflow orchestration must be serialized per workflow
+
+A workflow execution should have exactly one active mutator at a time when evaluating transitions and dispatching successor tasks.
+
+## Proposed Implementation Phases
+
+## Current Status
+
+As of the current implementation state:
+
+- Phase 1 is substantially implemented.
+- Phases 2, 3, 4, and 5 are implemented.
+
+Completed so far:
+
+- Atomic `requested -> scheduling` claim support was added in `ExecutionRepository`.
+- Scheduler state transitions for regular action dispatch were converted to conditional/CAS-style updates.
+- Redelivered `execution.requested` messages for stale `scheduling` rows are now retried/reclaimed instead of being silently acknowledged away.
+- Shared concurrency/FIFO coordination now uses durable PostgreSQL admission tables for action/group slot ownership and queued execution ordering.
+- `ExecutionQueueManager` now acts as a thin API-compatible facade over the DB-backed admission path when constructed with a pool.
+- Slot release, queued removal, and rollback/restore flows now operate against shared DB state rather than process-local memory.
+- `queue_stats` remains derived telemetry, but it is now refreshed transactionally from the shared admission state.
+- Workflow start is now idempotent at the parent workflow state level via `workflow_execution(execution)` uniqueness plus repository create-or-get behavior.
+- Workflow advancement now runs under a per-workflow PostgreSQL advisory lock, row-locks `workflow_execution` with `SELECT ... FOR UPDATE`, and performs serialized mutation inside an explicit SQL transaction.
+- Durable workflow child dispatch dedupe is now enforced with the `workflow_task_dispatch` coordination table and repository create-or-get helpers.
+- `execution` and `enforcement` were switched from Timescale hypertables back to normal PostgreSQL tables to remove HA/idempotency friction around foreign keys and unique constraints. `event` remains a hypertable, and history tables remain Timescale-backed.
+- Direct uniqueness/idempotency invariants were added for `enforcement(rule, event)`, top-level `execution(enforcement)`, and `inquiry(execution)`.
+- Event, enforcement, and inquiry handlers were updated to use create-or-get flows and conditional status transitions so duplicate delivery becomes safe.
+- Timeout and DLQ recovery loops now use conditional state transitions and only emit side effects when the guarded update succeeds.
+
+Partially complete / still open:
+
+- HA-focused integration and failure-injection coverage still needs to be expanded around the new invariants and recovery behavior.
+- The new migrations and DB-backed FIFO tests still need end-to-end validation against a real Postgres/Timescale environment.
+
+## Phase 1: Atomic Execution Claiming
+
+### Objective
+
+Ensure only one executor replica can claim a `requested` execution for scheduling.
+
+### Changes
+
+**`crates/common/src/repositories/execution.rs`**
+
+- Add a repository method for atomic status transition, for example:
+  - `claim_for_scheduling(id, expected_status, executor_id) -> Option<Execution>`
+  - or a more general compare-and-swap helper
+- Implement the claim as a single `UPDATE ... WHERE id = $1 AND status = 'requested' RETURNING ...`
+- Optionally persist the claiming replica identity in `execution.executor` for debugging and traceability
+
+**`crates/executor/src/scheduler.rs`**
+
+- Claim the execution before policy enforcement, worker selection, workflow start, or any side effects
+- Use `scheduling` as the claimed intermediate state
+- If the claim returns no row, treat the execution as already claimed/handled and acknowledge the message
+- Convert all later scheduler writes to conditional transitions from claimed state
+
+### Success Criteria
+
+- Two schedulers racing on the same execution cannot both dispatch it
+- Redelivered `execution.requested` messages become harmless no-ops after the first successful claim
+
+### Status
+
+Implemented for regular action scheduling, with additional stale-claim recovery for redelivered `execution.requested` messages. The remaining gap for this area is broader integration with the still-pending shared admission/queueing work in Phase 2.
+
+## Phase 2: Shared Concurrency Control and FIFO Queueing
+
+### Objective
+
+Replace the in-memory `ExecutionQueueManager` as the source of truth for concurrency slots and waiting order.
+
+### Changes
+
+**New schema**
+
+Add database-backed coordination tables, likely along these lines:
+
+- `execution_admission_slot`
+  - active slot ownership for an action/group key
+- `execution_admission_queue`
+  - ordered waiting executions for an action/group key
+
+Alternative naming is fine, but the design needs to support:
+
+- Action-level concurrency limits
+- Parameter-group concurrency keys
+- FIFO ordering within each action/group
+- Deterministic advancement when a slot is released
+
+**`crates/executor/src/policy_enforcer.rs`**
+
+- Replace `ExecutionQueueManager` slot acquisition with DB-backed admission logic
+- Keep existing policy semantics:
+  - `enqueue`
+  - `cancel`
+  - parameter-based concurrency grouping
+
+**`crates/executor/src/completion_listener.rs`**
+
+- Release the shared slot transactionally on completion
+- Select and wake the next queued execution in the same transaction
+- Republish only after the DB state is committed
+
+**`crates/executor/src/queue_manager.rs`**
+
+- Either remove it entirely or reduce it to a thin adapter over DB-backed coordination
+- Do not keep active slot ownership in process-local `DashMap`
+
+**`crates/common/src/repositories/queue_stats.rs`**
+
+- Keep `queue_stats` as derived telemetry only
+- Do not rely on it for correctness
+
+### Success Criteria
+
+- Completion processed by a different executor replica still releases the correct slot
+- FIFO ordering holds across multiple executor replicas
+- Restarting an executor does not lose queue ownership state
+
+### Status
+
+Implemented.
+
+Completed:
+
+- Shared admission state now lives in PostgreSQL via durable action/group queue rows and execution entry rows.
+- Action-level concurrency limits and parameter-group concurrency keys are enforced against that shared admission state.
+- FIFO ordering is determined by durable queued-entry order rather than process-local memory.
+- Completion-time slot release promotes the next queued execution inside the same DB transaction.
+- Rollback helpers can restore released slots or removed queued entries if republish/cleanup fails after the DB mutation.
+- `ExecutionQueueManager` remains as a facade for the existing scheduler/policy code paths, but it no longer acts as the correctness source of truth when running with a DB pool.
+
+## Phase 3: Workflow Start Idempotency and Serialized Advancement
+
+### Objective
+
+Ensure workflow orchestration is safe under concurrent replicas and duplicate completion messages.
+
+### Changes
+
+**Migration**
+
+Add a uniqueness constraint to guarantee one workflow state row per parent execution:
+
+```sql
+ALTER TABLE workflow_execution
+ADD CONSTRAINT uq_workflow_execution_execution UNIQUE (execution);
+```
+
+**`crates/executor/src/scheduler.rs`**
+
+- Change workflow start to be idempotent:
+  - either `INSERT ... ON CONFLICT ...`
+  - or claim parent execution first and only create workflow state once
+- When advancing a workflow:
+  - wrap read/decide/write logic in a transaction
+  - lock the `workflow_execution` row with `SELECT ... FOR UPDATE`
+  - or use advisory locks keyed by workflow execution id
+
+**Successor dispatch dedupe**
+
+Add a durable uniqueness guarantee for child task dispatch, for example:
+
+- one unique key for regular workflow tasks:
+  - `(workflow_execution_id, task_name, task_index IS NULL)`
+- one unique key for `with_items` children:
+  - `(workflow_execution_id, task_name, task_index)`
+
+This may be implemented with explicit columns or a dedupe table if indexing the current JSONB layout is awkward.
+
+**Repository support**
+
+- Add workflow repository helpers that support transactional locking and conditional updates
+- Avoid blind overwrite of `completed_tasks`, `failed_tasks`, and `variables` outside a serialized transaction
+
+### Success Criteria
+
+- Starting the same workflow twice cannot create two `workflow_execution` rows
+- Duplicate `execution.completed` delivery for a workflow child cannot create duplicate successor executions
+- Two executor replicas cannot concurrently mutate the same workflow state
+
+### Status
+
+Implemented.
+
+Completed:
+
+- `workflow_execution(execution)` uniqueness is part of the workflow schema and workflow start uses create-or-get semantics.
+- Workflow parent executions are claimed before orchestration starts.
+- Workflow advancement now runs under a per-workflow PostgreSQL advisory lock held on the same DB connection that performs the serialized advancement work.
+- The serialized workflow path is wrapped in an explicit SQL transaction.
+- `workflow_execution` is row-locked with `SELECT ... FOR UPDATE` before mutation.
+- Successor/child dispatch dedupe is enforced with the durable `workflow_task_dispatch` table keyed by `(workflow_execution, task_name, COALESCE(task_index, -1))`.
+- Child `ExecutionRequested` messages are staged and published only after the workflow transaction commits.
+
+## Phase 4: Idempotent Event, Enforcement, and Inquiry Handling
+
+### Objective
+
+Make duplicate delivery safe for all earlier and later executor-owned side effects.
+
+### Changes
+
+**Enforcement dedupe**
+
+Add a uniqueness rule so one event/rule pair produces at most one enforcement when `event` is present.
+
+Example:
+
+```sql
+CREATE UNIQUE INDEX uq_enforcement_rule_event
+ON enforcement(rule, event)
+WHERE event IS NOT NULL;
+```
+
+**`crates/executor/src/event_processor.rs`**
+
+- Use upsert-or-ignore semantics for enforcement creation
+- Treat uniqueness conflicts as idempotent success
+
+**`crates/executor/src/enforcement_processor.rs`**
+
+- Check current enforcement status before creating an execution
+- Add a durable relation that prevents an enforcement from creating more than one top-level execution
+- Options:
+  - unique partial index on `execution(enforcement)` for top-level executions
+  - or a separate coordination record
+
+**Inquiry dedupe**
+
+- Prevent duplicate inquiry creation per execution result/completion path
+- Add a unique invariant such as one active inquiry per execution, if that matches product semantics
+- Update completion handling to tolerate duplicate `execution.completed`
+
+### Success Criteria
+
+- Duplicate `event.created` does not create duplicate enforcements
+- Duplicate `enforcement.created` does not create duplicate executions
+- Duplicate completion handling does not create duplicate inquiries
+
+### Status
+
+Implemented.
+
+Completed:
+
+- `enforcement(rule, event)` uniqueness is enforced directly with a partial unique index when both keys are present.
+- Top-level execution creation is deduped with a unique invariant on `execution(enforcement)` where `parent IS NULL`.
+- Inquiry creation is deduped with a unique invariant on `inquiry(execution)`.
+- `event_processor` now uses create-or-get enforcement handling and only republishes when the persisted enforcement still needs processing.
+- `enforcement_processor` now skips duplicate non-`created` enforcements, creates or reuses the top-level execution, and conditionally resolves enforcement state.
+- `inquiry_handler` now uses create-or-get inquiry handling and only emits `InquiryCreated` when the inquiry was actually created.
+
+## Phase 5: Safe Recovery Loops
+
+### Objective
+
+Make timeout and DLQ processing safe under races and multiple replicas.
+
+### Changes
+
+**`crates/executor/src/timeout_monitor.rs`**
+
+- Replace unconditional updates with conditional state transitions:
+  - `UPDATE execution SET ... WHERE id = $1 AND status = 'scheduled' ... RETURNING ...`
+- Only publish completion side effects when a row was actually updated
+- Consider including `updated < cutoff` in the same update statement
+
+**`crates/executor/src/dead_letter_handler.rs`**
+
+- Change failure transition to conditional update based on current state
+- Do not overwrite executions that have already moved to `running` or terminal state
+- Only emit side effects when the row transition succeeded
+
+**`crates/executor/src/service.rs`**
+
+- It is acceptable for these loops to run on every replica once updates are conditional
+- Optional future optimization: leader election for janitor loops to reduce duplicate scans and log noise
+
+### Success Criteria
+
+- Timeout monitor cannot fail an execution that has already moved to `running`
+- DLQ handler cannot overwrite newer state
+- Running multiple timeout monitors produces no conflicting state transitions
+
+### Status
+
+Implemented.
+
+Completed:
+
+- Timeout failure now uses a conditional transition that only succeeds when the execution is still `scheduled` and still older than the timeout cutoff.
+- Timeout-driven completion side effects are only published when that guarded update succeeds.
+- DLQ handling now treats messages as stale unless the execution is still exactly `scheduled`.
+- DLQ failure transitions now use conditional status updates and no longer overwrite newer `running` or terminal state.
+
+## Testing Plan
+
+Add focused HA tests after the repository and scheduler primitives are in place.
+
+### Repository tests
+
+- Compare-and-swap execution claim succeeds exactly once
+- Conditional timeout/DLQ transition updates exactly one row or zero rows as expected
+- Workflow uniqueness constraint prevents duplicate workflow state rows
+
+### Executor integration tests
+
+- Two scheduler instances processing the same `execution.requested` message only dispatch once
+- Completion consumed by a different executor replica still advances the shared queue
+- Duplicate workflow child completion does not create duplicate successor tasks
+- Duplicate `event.created` and `enforcement.created` messages do not create duplicate downstream records
+
+### Failure-injection tests
+
+- Executor crashes after claiming but before publish
+- Executor crashes after slot release but before republish
+- Duplicate `execution.completed` delivery after successful workflow advancement
+
+## Recommended Execution Order for Next Session
+
+1. Add more HA-focused integration tests for duplicate delivery, cross-replica completion, and recovery rollback paths
+2. Add failure-injection tests for crash/replay scenarios around `scheduling` reclaim, workflow advancement, and post-commit publish paths
+3. Validate the new migrations and DB-backed FIFO behavior end-to-end against a real Postgres/Timescale environment
+4. Consider a small follow-up cleanup pass to reduce or remove the in-memory fallback code in `ExecutionQueueManager` once the DB path is fully baked
+
+## Expected Outcome
+
+After this plan is implemented, the executor should be able to scale horizontally without relying on singleton behavior. Multiple executor replicas should be able to process work concurrently while preserving:
+
+- exactly-once scheduling semantics at the execution state level
+- shared concurrency limits and FIFO behavior
+- correct workflow orchestration
+- safe replay handling
+- safe recovery behavior during failures and redelivery
+
+At the current state, the core executor HA phases are implemented. The remaining work is confidence-building: failure-injection coverage, multi-replica integration testing, and end-to-end migration validation in a live database environment.

migrations/20250101000004_trigger_sensor_event_rule.sql

@@ -201,6 +201,9 @@ CREATE INDEX idx_enforcement_rule_status ON enforcement(rule, status);
 CREATE INDEX idx_enforcement_event_status ON enforcement(event, status);
 CREATE INDEX idx_enforcement_payload_gin ON enforcement USING GIN (payload);
 CREATE INDEX idx_enforcement_conditions_gin ON enforcement USING GIN (conditions);
+CREATE UNIQUE INDEX uq_enforcement_rule_event
+    ON enforcement (rule, event)
+    WHERE rule IS NOT NULL AND event IS NOT NULL;
 
 -- Comments
 COMMENT ON TABLE enforcement IS 'Enforcements represent rule triggering by events';

migrations/20250101000005_execution_and_operations.sql

@@ -4,13 +4,8 @@
 --              Consolidates former migrations: 000006 (execution_system), 000008
 --              (worker_notification), 000014 (worker_table), and 20260209 (phase3).
 --
---              NOTE: The execution table is converted to a TimescaleDB hypertable in
---              migration 000009. Hypertables cannot be the target of FK constraints,
---              so columns referencing execution (inquiry.execution, workflow_execution.execution)
---              are plain BIGINT with no FK. Similarly, columns ON the execution table that
---              would self-reference or reference other hypertables (parent, enforcement,
---              original_execution) are plain BIGINT. The action and executor FKs are also
---              omitted since they would need to be dropped during hypertable conversion.
+--              NOTE: `execution` remains a regular PostgreSQL table. Time-series
+--              audit and analytics are handled by `execution_history`.
 -- Version: 20250101000005
 
 -- ============================================================================
@@ -19,27 +14,27 @@
 
 CREATE TABLE execution (
     id BIGSERIAL PRIMARY KEY,
-    action BIGINT,          -- references action(id); no FK because execution becomes a hypertable
+    action BIGINT,
     action_ref TEXT NOT NULL,
     config JSONB,
     env_vars JSONB,
-    parent BIGINT,          -- self-reference; no FK because execution becomes a hypertable
-    enforcement BIGINT,     -- references enforcement(id); no FK (both are hypertables)
-    executor BIGINT,        -- references identity(id); no FK because execution becomes a hypertable
-    worker BIGINT,          -- references worker(id); no FK because execution becomes a hypertable
+    parent BIGINT,
+    enforcement BIGINT,
+    executor BIGINT,
+    worker BIGINT,
     status execution_status_enum NOT NULL DEFAULT 'requested',
     result JSONB,
     started_at TIMESTAMPTZ,         -- set when execution transitions to 'running'
     created TIMESTAMPTZ NOT NULL DEFAULT NOW(),
     is_workflow BOOLEAN DEFAULT false NOT NULL,
-    workflow_def BIGINT,    -- references workflow_definition(id); no FK because execution becomes a hypertable
+    workflow_def BIGINT,
     workflow_task JSONB,
 
     -- Retry tracking (baked in from phase 3)
     retry_count INTEGER NOT NULL DEFAULT 0,
     max_retries INTEGER,
     retry_reason TEXT,
-    original_execution BIGINT, -- self-reference; no FK because execution becomes a hypertable
+    original_execution BIGINT,
 
     updated TIMESTAMPTZ NOT NULL DEFAULT NOW()
 );
@@ -64,6 +59,11 @@ CREATE INDEX idx_execution_result_gin ON execution USING GIN (result);
 CREATE INDEX idx_execution_env_vars_gin ON execution USING GIN (env_vars);
 CREATE INDEX idx_execution_original_execution ON execution(original_execution) WHERE original_execution IS NOT NULL;
 CREATE INDEX idx_execution_status_retry ON execution(status, retry_count) WHERE status = 'failed' AND retry_count < COALESCE(max_retries, 0);
+CREATE UNIQUE INDEX uq_execution_top_level_enforcement
+    ON execution (enforcement)
+    WHERE enforcement IS NOT NULL
+      AND parent IS NULL
+      AND (config IS NULL OR NOT (config ? 'retry_of'));
 
 -- Trigger
 CREATE TRIGGER update_execution_updated
@@ -77,10 +77,10 @@ COMMENT ON COLUMN execution.action IS 'Action being executed (may be null if act
 COMMENT ON COLUMN execution.action_ref IS 'Action reference (preserved even if action deleted)';
 COMMENT ON COLUMN execution.config IS 'Snapshot of action configuration at execution time';
 COMMENT ON COLUMN execution.env_vars IS 'Environment variables for this execution as key-value pairs (string -> string). These are set in the execution environment and are separate from action parameters. Used for execution context, configuration, and non-sensitive metadata.';
-COMMENT ON COLUMN execution.parent IS 'Parent execution ID for workflow hierarchies (no FK — execution is a hypertable)';
-COMMENT ON COLUMN execution.enforcement IS 'Enforcement that triggered this execution (no FK — both are hypertables)';
-COMMENT ON COLUMN execution.executor IS 'Identity that initiated the execution (no FK — execution is a hypertable)';
-COMMENT ON COLUMN execution.worker IS 'Assigned worker handling this execution (no FK — execution is a hypertable)';
+COMMENT ON COLUMN execution.parent IS 'Parent execution ID for workflow hierarchies';
+COMMENT ON COLUMN execution.enforcement IS 'Enforcement that triggered this execution';
+COMMENT ON COLUMN execution.executor IS 'Identity that initiated the execution';
+COMMENT ON COLUMN execution.worker IS 'Assigned worker handling this execution';
 COMMENT ON COLUMN execution.status IS 'Current execution lifecycle status';
 COMMENT ON COLUMN execution.result IS 'Execution output/results';
 COMMENT ON COLUMN execution.retry_count IS 'Current retry attempt number (0 = first attempt, 1 = first retry, etc.)';
@@ -96,7 +96,7 @@ COMMENT ON COLUMN execution.original_execution IS 'ID of the original execution
 
 CREATE TABLE inquiry (
     id BIGSERIAL PRIMARY KEY,
-    execution BIGINT NOT NULL, -- references execution(id); no FK because execution is a hypertable
+    execution BIGINT NOT NULL,
     prompt TEXT NOT NULL,
     response_schema JSONB,
     assigned_to BIGINT REFERENCES identity(id) ON DELETE SET NULL,
@@ -109,7 +109,7 @@ CREATE TABLE inquiry (
 );
 
 -- Indexes
-CREATE INDEX idx_inquiry_execution ON inquiry(execution);
+CREATE UNIQUE INDEX uq_inquiry_execution ON inquiry(execution) WHERE execution IS NOT NULL;
 CREATE INDEX idx_inquiry_assigned_to ON inquiry(assigned_to);
 CREATE INDEX idx_inquiry_status ON inquiry(status);
 CREATE INDEX idx_inquiry_timeout_at ON inquiry(timeout_at) WHERE timeout_at IS NOT NULL;
@@ -127,7 +127,31 @@ CREATE TRIGGER update_inquiry_updated
 
 -- Comments
 COMMENT ON TABLE inquiry IS 'Inquiries enable human-in-the-loop workflows with async user interactions';
-COMMENT ON COLUMN inquiry.execution IS 'Execution that is waiting on this inquiry (no FK — execution is a hypertable)';
+COMMENT ON COLUMN inquiry.execution IS 'Execution that is waiting on this inquiry';
+
+ALTER TABLE execution
+    ADD CONSTRAINT execution_action_fkey
+    FOREIGN KEY (action) REFERENCES action(id) ON DELETE SET NULL;
+
+ALTER TABLE execution
+    ADD CONSTRAINT execution_parent_fkey
+    FOREIGN KEY (parent) REFERENCES execution(id) ON DELETE SET NULL;
+
+ALTER TABLE execution
+    ADD CONSTRAINT execution_original_execution_fkey
+    FOREIGN KEY (original_execution) REFERENCES execution(id) ON DELETE SET NULL;
+
+ALTER TABLE execution
+    ADD CONSTRAINT execution_enforcement_fkey
+    FOREIGN KEY (enforcement) REFERENCES enforcement(id) ON DELETE SET NULL;
+
+ALTER TABLE execution
+    ADD CONSTRAINT execution_executor_fkey
+    FOREIGN KEY (executor) REFERENCES identity(id) ON DELETE SET NULL;
+
+ALTER TABLE inquiry
+    ADD CONSTRAINT inquiry_execution_fkey
+    FOREIGN KEY (execution) REFERENCES execution(id) ON DELETE CASCADE;
 COMMENT ON COLUMN inquiry.prompt IS 'Question or prompt text for the user';
 COMMENT ON COLUMN inquiry.response_schema IS 'JSON schema defining expected response format';
 COMMENT ON COLUMN inquiry.assigned_to IS 'Identity who should respond to this inquiry';
@@ -261,6 +285,10 @@ COMMENT ON COLUMN worker.capabilities IS 'Worker capabilities (e.g., max_concurr
 COMMENT ON COLUMN worker.meta IS 'Additional worker metadata';
 COMMENT ON COLUMN worker.last_heartbeat IS 'Timestamp of last heartbeat from worker';
 
+ALTER TABLE execution
+    ADD CONSTRAINT execution_worker_fkey
+    FOREIGN KEY (worker) REFERENCES worker(id) ON DELETE SET NULL;
+
 -- ============================================================================
 -- NOTIFICATION TABLE
 -- ============================================================================

migrations/20250101000006_workflow_system.sql

@@ -1,13 +1,11 @@
 -- Migration: Workflow System
--- Description: Creates workflow_definition and workflow_execution tables
+-- Description: Creates workflow_definition, workflow_execution, and
+--              workflow_task_dispatch tables
 --              (workflow_task_execution consolidated into execution.workflow_task JSONB)
 --
---              NOTE: The execution table is converted to a TimescaleDB hypertable in
---              migration 000009. Hypertables cannot be the target of FK constraints,
---              so workflow_execution.execution is a plain BIGINT with no FK.
---              execution.workflow_def also has no FK (added as plain BIGINT in 000005)
---              since execution is a hypertable and FKs from hypertables are only
---              supported for simple cases — we omit it for consistency.
+--              NOTE: `execution` remains a regular PostgreSQL table, so
+--              workflow_execution.execution, workflow_task_dispatch.execution_id,
+--              and execution.workflow_def use normal foreign keys.
 -- Version: 20250101000006
 
 -- ============================================================================
@@ -54,7 +52,7 @@ COMMENT ON COLUMN workflow_definition.out_schema IS 'JSON schema for workflow ou
 
 CREATE TABLE workflow_execution (
     id BIGSERIAL PRIMARY KEY,
-    execution BIGINT NOT NULL, -- references execution(id); no FK because execution is a hypertable
+    execution BIGINT NOT NULL REFERENCES execution(id) ON DELETE CASCADE,
     workflow_def BIGINT NOT NULL REFERENCES workflow_definition(id) ON DELETE CASCADE,
     current_tasks TEXT[] DEFAULT '{}',
     completed_tasks TEXT[] DEFAULT '{}',
@@ -67,11 +65,11 @@ CREATE TABLE workflow_execution (
     paused BOOLEAN DEFAULT false NOT NULL,
     pause_reason TEXT,
     created TIMESTAMPTZ DEFAULT NOW() NOT NULL,
-    updated TIMESTAMPTZ DEFAULT NOW() NOT NULL
+    updated TIMESTAMPTZ DEFAULT NOW() NOT NULL,
+    CONSTRAINT uq_workflow_execution_execution UNIQUE (execution)
 );
 
 -- Indexes
-CREATE INDEX idx_workflow_exec_execution ON workflow_execution(execution);
 CREATE INDEX idx_workflow_exec_workflow_def ON workflow_execution(workflow_def);
 CREATE INDEX idx_workflow_exec_status ON workflow_execution(status);
 CREATE INDEX idx_workflow_exec_paused ON workflow_execution(paused) WHERE paused = true;
@@ -83,12 +81,51 @@ CREATE TRIGGER update_workflow_execution_updated
     EXECUTE FUNCTION update_updated_column();
 
 -- Comments
-COMMENT ON TABLE workflow_execution IS 'Runtime state tracking for workflow executions. execution column has no FK — execution is a hypertable.';
+COMMENT ON TABLE workflow_execution IS 'Runtime state tracking for workflow executions.';
 COMMENT ON COLUMN workflow_execution.variables IS 'Workflow-scoped variables, updated via publish directives';
 COMMENT ON COLUMN workflow_execution.task_graph IS 'Execution graph with dependencies and transitions';
 COMMENT ON COLUMN workflow_execution.current_tasks IS 'Array of task names currently executing';
 COMMENT ON COLUMN workflow_execution.paused IS 'True if workflow execution is paused (can be resumed)';
 
+-- ============================================================================
+-- WORKFLOW TASK DISPATCH TABLE
+-- ============================================================================
+
+CREATE TABLE workflow_task_dispatch (
+    id BIGSERIAL PRIMARY KEY,
+    workflow_execution BIGINT NOT NULL REFERENCES workflow_execution(id) ON DELETE CASCADE,
+    task_name TEXT NOT NULL,
+    task_index INT,
+    execution_id BIGINT,
+    created TIMESTAMPTZ DEFAULT NOW() NOT NULL,
+    updated TIMESTAMPTZ DEFAULT NOW() NOT NULL
+);
+
+CREATE UNIQUE INDEX uq_workflow_task_dispatch_identity
+    ON workflow_task_dispatch (
+        workflow_execution,
+        task_name,
+        COALESCE(task_index, -1)
+    );
+
+CREATE INDEX idx_workflow_task_dispatch_execution_id
+    ON workflow_task_dispatch (execution_id)
+    WHERE execution_id IS NOT NULL;
+
+CREATE TRIGGER update_workflow_task_dispatch_updated
+    BEFORE UPDATE ON workflow_task_dispatch
+    FOR EACH ROW
+    EXECUTE FUNCTION update_updated_column();
+
+COMMENT ON TABLE workflow_task_dispatch IS
+    'Durable dedupe/ownership records for workflow child execution dispatch';
+COMMENT ON COLUMN workflow_task_dispatch.execution_id IS
+    'Associated execution.id';
+
+ALTER TABLE workflow_task_dispatch
+    ADD CONSTRAINT workflow_task_dispatch_execution_id_fkey
+    FOREIGN KEY (execution_id) REFERENCES execution(id) ON DELETE CASCADE;
+
 -- ============================================================================
 -- MODIFY ACTION TABLE - Add Workflow Support
 -- ============================================================================
@@ -100,9 +137,9 @@ CREATE INDEX idx_action_workflow_def ON action(workflow_def);
 
 COMMENT ON COLUMN action.workflow_def IS 'Reference to workflow definition (non-null means this action is a workflow)';
 
--- NOTE: execution.workflow_def has no FK constraint because execution is a
--- TimescaleDB hypertable (converted in migration 000009). The column was
--- created as a plain BIGINT in migration 000005.
+ALTER TABLE execution
+    ADD CONSTRAINT execution_workflow_def_fkey
+    FOREIGN KEY (workflow_def) REFERENCES workflow_definition(id) ON DELETE SET NULL;
 
 -- ============================================================================
 -- WORKFLOW VIEWS

migrations/20250101000007_supporting_systems.sql

@@ -1,6 +1,6 @@
 -- Migration: Supporting Systems
--- Description: Creates keys, artifacts, queue_stats, pack_environment, pack_testing,
---              and webhook function tables.
+-- Description: Creates keys, artifacts, queue_stats, execution_admission,
+--              pack_environment, pack_testing, and webhook function tables.
 --              Consolidates former migrations: 000009 (keys_artifacts), 000010 (webhook_system),
 --              000011 (pack_environments), and 000012 (pack_testing).
 -- Version: 20250101000007
@@ -206,6 +206,76 @@ COMMENT ON COLUMN queue_stats.total_enqueued IS 'Total executions enqueued since
 COMMENT ON COLUMN queue_stats.total_completed IS 'Total executions completed since queue creation';
 COMMENT ON COLUMN queue_stats.last_updated IS 'Timestamp of last statistics update';
 
+-- ============================================================================
+-- EXECUTION ADMISSION TABLES
+-- ============================================================================
+
+CREATE TABLE execution_admission_state (
+    id BIGSERIAL PRIMARY KEY,
+    action_id BIGINT NOT NULL REFERENCES action(id) ON DELETE CASCADE,
+    group_key TEXT,
+    group_key_normalized TEXT GENERATED ALWAYS AS (COALESCE(group_key, '')) STORED,
+    max_concurrent INTEGER NOT NULL,
+    next_queue_order BIGINT NOT NULL DEFAULT 1,
+    total_enqueued BIGINT NOT NULL DEFAULT 0,
+    total_completed BIGINT NOT NULL DEFAULT 0,
+    created TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    updated TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    CONSTRAINT uq_execution_admission_state_identity
+        UNIQUE (action_id, group_key_normalized)
+);
+
+CREATE TABLE execution_admission_entry (
+    id BIGSERIAL PRIMARY KEY,
+    state_id BIGINT NOT NULL REFERENCES execution_admission_state(id) ON DELETE CASCADE,
+    execution_id BIGINT NOT NULL UNIQUE REFERENCES execution(id) ON DELETE CASCADE,
+    status TEXT NOT NULL CHECK (status IN ('active', 'queued')),
+    queue_order BIGINT NOT NULL,
+    enqueued_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    activated_at TIMESTAMPTZ,
+    created TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    updated TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+
+CREATE INDEX idx_execution_admission_state_action
+    ON execution_admission_state (action_id);
+
+CREATE INDEX idx_execution_admission_entry_state_status_queue
+    ON execution_admission_entry (state_id, status, queue_order);
+
+CREATE INDEX idx_execution_admission_entry_execution
+    ON execution_admission_entry (execution_id);
+
+CREATE TRIGGER update_execution_admission_state_updated
+    BEFORE UPDATE ON execution_admission_state
+    FOR EACH ROW
+    EXECUTE FUNCTION update_updated_column();
+
+CREATE TRIGGER update_execution_admission_entry_updated
+    BEFORE UPDATE ON execution_admission_entry
+    FOR EACH ROW
+    EXECUTE FUNCTION update_updated_column();
+
+COMMENT ON TABLE execution_admission_state IS
+    'Shared admission state per action/group for executor concurrency and FIFO coordination';
+COMMENT ON COLUMN execution_admission_state.group_key IS
+    'Optional parameter-derived concurrency grouping key';
+COMMENT ON COLUMN execution_admission_state.max_concurrent IS
+    'Current concurrency limit for this action/group queue';
+COMMENT ON COLUMN execution_admission_state.next_queue_order IS
+    'Monotonic sequence used to preserve exact FIFO order for queued executions';
+COMMENT ON COLUMN execution_admission_state.total_enqueued IS
+    'Cumulative number of executions admitted into this queue';
+COMMENT ON COLUMN execution_admission_state.total_completed IS
+    'Cumulative number of active executions released from this queue';
+
+COMMENT ON TABLE execution_admission_entry IS
+    'Active slot ownership and queued executions for shared admission control';
+COMMENT ON COLUMN execution_admission_entry.status IS
+    'active rows own a concurrency slot; queued rows wait in FIFO order';
+COMMENT ON COLUMN execution_admission_entry.queue_order IS
+    'Durable FIFO position within an action/group queue';
+
 -- ============================================================================
 -- PACK ENVIRONMENT TABLE
 -- ============================================================================

migrations/20250101000009_timescaledb_history.sql

@@ -143,52 +143,8 @@ SELECT create_hypertable('event', 'created',
 
 COMMENT ON TABLE event IS 'Events are instances of triggers firing (TimescaleDB hypertable partitioned on created)';
 
--- ============================================================================
--- CONVERT ENFORCEMENT TABLE TO HYPERTABLE
--- ============================================================================
--- Enforcements are created and then updated exactly once (status changes from
--- `created` to `processed` or `disabled` within ~1 second). This single update
--- happens well before the 7-day compression window, so UPDATE on uncompressed
--- chunks works without issues.
---
--- No FK constraints reference enforcement(id) — execution.enforcement was
--- created as a plain BIGINT in migration 000005.
--- ----------------------------------------------------------------------------
-
-ALTER TABLE enforcement DROP CONSTRAINT enforcement_pkey;
-ALTER TABLE enforcement ADD PRIMARY KEY (id, created);
-
-SELECT create_hypertable('enforcement', 'created',
-    chunk_time_interval => INTERVAL '1 day',
-    migrate_data        => true);
-
-COMMENT ON TABLE enforcement IS 'Enforcements represent rule triggering by events (TimescaleDB hypertable partitioned on created)';
-
--- ============================================================================
--- CONVERT EXECUTION TABLE TO HYPERTABLE
--- ============================================================================
--- Executions are updated ~4 times during their lifecycle (requested → scheduled
--- → running → completed/failed), completing within at most ~1 day — well before
--- the 7-day compression window. The `updated` column and its BEFORE UPDATE
--- trigger are preserved (used by timeout monitor and UI).
---
--- No FK constraints reference execution(id) — inquiry.execution,
--- workflow_execution.execution, execution.parent, and execution.original_execution
--- were all created as plain BIGINT columns in migrations 000005 and 000006.
---
--- The existing execution_history hypertable and its trigger are preserved —
--- they track field-level diffs of each update, which remains valuable for
--- a mutable table.
--- ----------------------------------------------------------------------------
-
-ALTER TABLE execution DROP CONSTRAINT execution_pkey;
-ALTER TABLE execution ADD PRIMARY KEY (id, created);
-
-SELECT create_hypertable('execution', 'created',
-    chunk_time_interval => INTERVAL '1 day',
-    migrate_data        => true);
-
-COMMENT ON TABLE execution IS 'Executions represent action runs with workflow support (TimescaleDB hypertable partitioned on created). Updated ~4 times during lifecycle, completing within ~1 day (well before 7-day compression window).';
+COMMENT ON TABLE enforcement IS 'Enforcements represent rule triggering by events';
+COMMENT ON TABLE execution IS 'Executions represent action runs with workflow support. History and analytics are stored in execution_history.';
 
 -- ============================================================================
 -- TRIGGER FUNCTIONS
@@ -410,22 +366,6 @@ ALTER TABLE event SET (
 );
 SELECT add_compression_policy('event', INTERVAL '7 days');
 
--- Enforcement table (hypertable)
-ALTER TABLE enforcement SET (
-    timescaledb.compress,
-    timescaledb.compress_segmentby = 'rule_ref',
-    timescaledb.compress_orderby = 'created DESC'
-);
-SELECT add_compression_policy('enforcement', INTERVAL '7 days');
-
--- Execution table (hypertable)
-ALTER TABLE execution SET (
-    timescaledb.compress,
-    timescaledb.compress_segmentby = 'action_ref',
-    timescaledb.compress_orderby = 'created DESC'
-);
-SELECT add_compression_policy('execution', INTERVAL '7 days');
-
 -- ============================================================================
 -- RETENTION POLICIES
 -- ============================================================================
@@ -433,8 +373,6 @@ SELECT add_compression_policy('execution', INTERVAL '7 days');
 SELECT add_retention_policy('execution_history', INTERVAL '90 days');
 SELECT add_retention_policy('worker_history', INTERVAL '180 days');
 SELECT add_retention_policy('event', INTERVAL '90 days');
-SELECT add_retention_policy('enforcement', INTERVAL '90 days');
-SELECT add_retention_policy('execution', INTERVAL '90 days');
 
 -- ============================================================================
 -- CONTINUOUS AGGREGATES
@@ -449,6 +387,8 @@ DROP MATERIALIZED VIEW IF EXISTS event_volume_hourly CASCADE;
 DROP MATERIALIZED VIEW IF EXISTS worker_status_hourly CASCADE;
 DROP MATERIALIZED VIEW IF EXISTS enforcement_volume_hourly CASCADE;
 DROP MATERIALIZED VIEW IF EXISTS execution_volume_hourly CASCADE;
+DROP VIEW IF EXISTS enforcement_volume_hourly CASCADE;
+DROP VIEW IF EXISTS execution_volume_hourly CASCADE;
 
 -- ----------------------------------------------------------------------------
 -- execution_status_hourly
@@ -553,49 +493,35 @@ SELECT add_continuous_aggregate_policy('worker_status_hourly',
 --       instead of a separate enforcement_history table.
 -- ----------------------------------------------------------------------------
 
-CREATE MATERIALIZED VIEW enforcement_volume_hourly
-WITH (timescaledb.continuous) AS
+CREATE VIEW enforcement_volume_hourly AS
 SELECT
-    time_bucket('1 hour', created) AS bucket,
+    date_trunc('hour', created) AS bucket,
     rule_ref,
     COUNT(*) AS enforcement_count
 FROM enforcement
 GROUP BY bucket, rule_ref
-WITH NO DATA;
-
-SELECT add_continuous_aggregate_policy('enforcement_volume_hourly',
-    start_offset    => INTERVAL '7 days',
-    end_offset      => INTERVAL '1 hour',
-    schedule_interval => INTERVAL '30 minutes'
-);
+;
 
 -- ----------------------------------------------------------------------------
 -- execution_volume_hourly
 -- Tracks execution creation volume per hour by action_ref and status.
--- This queries the execution hypertable directly (like event_volume_hourly
--- queries the event table). Complements the existing execution_status_hourly
--- and execution_throughput_hourly aggregates which query execution_history.
+-- This queries the execution table directly. Complements the existing
+-- execution_status_hourly and execution_throughput_hourly aggregates which
+-- query execution_history.
 --
 -- Use case: direct execution volume monitoring without relying on the history
 -- trigger (belt-and-suspenders, plus captures the initial status at creation).
 -- ----------------------------------------------------------------------------
 
-CREATE MATERIALIZED VIEW execution_volume_hourly
-WITH (timescaledb.continuous) AS
+CREATE VIEW execution_volume_hourly AS
 SELECT
-    time_bucket('1 hour', created) AS bucket,
+    date_trunc('hour', created) AS bucket,
     action_ref,
     status AS initial_status,
     COUNT(*) AS execution_count
 FROM execution
 GROUP BY bucket, action_ref, status
-WITH NO DATA;
-
-SELECT add_continuous_aggregate_policy('execution_volume_hourly',
-    start_offset    => INTERVAL '7 days',
-    end_offset      => INTERVAL '1 hour',
-    schedule_interval => INTERVAL '30 minutes'
-);
+;
 
 -- ============================================================================
 -- INITIAL REFRESH NOTE

migrations/20250101000010_artifact_content.sql

@@ -26,7 +26,7 @@ ALTER TABLE artifact ADD COLUMN IF NOT EXISTS content_type TEXT;
 -- Total size in bytes of the latest version's content (NULL for progress artifacts)
 ALTER TABLE artifact ADD COLUMN IF NOT EXISTS size_bytes BIGINT;
 
--- Execution that produced/owns this artifact (plain BIGINT, no FK — execution is a hypertable)
+-- Execution that produced/owns this artifact (plain BIGINT, no FK by design)
 ALTER TABLE artifact ADD COLUMN IF NOT EXISTS execution BIGINT;
 
 -- Structured data for progress-type artifacts and small structured payloads.
@@ -52,7 +52,7 @@ COMMENT ON COLUMN artifact.name IS 'Human-readable artifact name';
 COMMENT ON COLUMN artifact.description IS 'Optional description of the artifact';
 COMMENT ON COLUMN artifact.content_type IS 'MIME content type (e.g. application/json, text/plain)';
 COMMENT ON COLUMN artifact.size_bytes IS 'Size of latest version content in bytes';
-COMMENT ON COLUMN artifact.execution IS 'Execution that produced this artifact (no FK — execution is a hypertable)';
+COMMENT ON COLUMN artifact.execution IS 'Execution that produced this artifact (no FK by design)';
 COMMENT ON COLUMN artifact.data IS 'Structured JSONB data for progress artifacts or metadata';
 COMMENT ON COLUMN artifact.visibility IS 'Access visibility: public (all users) or private (scope/owner-restricted)';