**Symptom:** Webhook handler stops processing after a few hundred messages, no error logs. **Expected:** Handler keeps consuming messages indefinitely (source: continuous webhook stream from the upstream provider). **Likely location:** handle_webhook() in webhooks.py, or upstream of it in the request pipeline. **Candidate causes:** 1. **Database connection pool exhaustion.** The handler opens a connection per message and never returns it. After the pool fills, new requests hang silently. Falsifying test: run the handler with SQLAlchemy.engine.pool logging enabled. If pool usage climbs to the limit and stays there, this is the cause. 2. **Background task queue saturating.** If background_tasks.add_task is being used and tasks aren't completing, the queue fills until the worker process degrades. Falsifying test: log queue size on every request. If it climbs without dropping, this is the cause. 3. **Memory leak from accumulated state.** Some object is being held in module-level state (a list, a dict) that grows without bound. Falsifying test: capture tracemalloc snapshots at message 100 and message 400. If the heap diff shows a single growing structure, this is the cause. **Recommended fix (assuming cause #1 is right):** Use the session-per-request pattern via a FastAPI dependency, not a session created inside the handler. def get_db(): db = SessionLocal() try: yield db finally: db.close() @app.post("/webhook") async def handle_webhook(payload: dict, db: Session = Depends(get_db)): ... **If cause #1 is wrong:** Run the falsifying tests for causes #2 and #3 in parallel. Cause #3 is easy to rule out quickly with tracemalloc; cause #2 needs you to inspect whether background_tasks is used.