Add event-driven reporting and durable failure recovery hooks

scripts/worker.py

@@ -260,6 +260,7 @@ def work(cfg, once=False):
         if state.get('inflight'):
             raise RuntimeError('Previous execution was interrupted; inspect its thread and side effects before explicit retry')
         response = client.call('/api/worker/claim', {'worker_id': worker_id, 'session_id': session_id, 'lease_seconds': 120})
+        atomic_json(state_path.parent/'worker-health.json',{'status':'ok','updated':time.time()})
         if response.get('needs_reconciliation'):
             raise RuntimeError('Server session requires reconciliation; no message will be executed')
         message = response.get('message')
@@ -269,6 +270,7 @@ def work(cfg, once=False):
             state['inflight'] = {**lease, 'started_at': datetime.now(timezone.utc).isoformat()}
             atomic_json(state_path, state)
             def renew():
+                atomic_json(state_path.parent/'worker-health.json',{'status':'ok','updated':time.time()})
                 try:
                     client.call('/api/worker/renew', {**lease, 'lease_seconds': 120})
                 except APIError as exc: