BAI-3: harden live readiness diagnostics

Co-authored-by: multica-agent <github@multica.ai>

app/backend/jobs/repository.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 import json
 from dataclasses import dataclass
-from datetime import datetime
+from datetime import datetime, timedelta
 from typing import Any
 
 from database import ReviewDatabase
@@ -27,9 +27,37 @@ class SQLiteJobRunRepository:
     def start(
         self, job_id: str, idempotency_key: str, output_version: str,
         metadata: dict[str, Any] | None = None,
-    ) -> int:
-        now = datetime.now().astimezone().isoformat(timespec="seconds")
+        stale_after_seconds: int = 0,
+    ) -> int | None:
+        current = datetime.now().astimezone()
+        now = current.isoformat(timespec="seconds")
+        stale_before = (
+            current - timedelta(seconds=max(1, int(stale_after_seconds or 1)))
+        ).isoformat(timespec="seconds")
         with self.database.connect() as connection:
+            # The process-local runner lock cannot protect two server processes.
+            connection.execute("BEGIN IMMEDIATE")
+            connection.execute(
+                """
+                UPDATE job_runs
+                SET status = 'failed', finished_at = ?, error_code = 'StaleRun',
+                    message = 'Previous run exceeded its execution window.'
+                WHERE job_id = ? AND idempotency_key = ? AND status = 'running'
+                  AND started_at < ?
+                """,
+                (now, job_id, idempotency_key, stale_before),
+            )
+            claimed = connection.execute(
+                """
+                SELECT 1 FROM job_runs
+                WHERE job_id = ? AND idempotency_key = ?
+                  AND status IN ('running', 'success')
+                LIMIT 1
+                """,
+                (job_id, idempotency_key),
+            ).fetchone()
+            if claimed:
+                return None
             row = connection.execute(
                 """
                 SELECT COALESCE(MAX(attempt), 0) + 1 AS attempt FROM job_runs