# 2026-06-18

- 2026-06-18 heartbeat wake: gateway runtime is healthy with an ok connectivity probe, session count is 10, and no user-facing attention is needed at 03:43 PDT.
- 2026-06-18 14:55 PDT: Sentinel alert reported canonical `agent:main:main` missing from session store, but live checks showed the gateway healthy, `openclaw sessions` count at 10, and `agent:main:main` present/running in `session_status` and `sessions_list`. Treat as transient bookkeeping drift unless it repeats.
- 2026-06-18 09:02 PDT: Braden reported extreme Codex/OpenClaw drain. Investigation found `ai.openclaw.node` in a launchd restart loop (`launchctl print` showed `state = spawn scheduled`, `runs = 126931`). The job was repeatedly attempting a localhost gateway role upgrade (`operator -> node`) and spawning hot `openclaw-hooks` children while flooding `/tmp/openclaw/openclaw-2026-06-18.log` with `role upgrade pending approval` errors. Mitigation applied: `launchctl disable gui/501/ai.openclaw.node` and `launchctl bootout gui/501/ai.openclaw.node`, followed by terminating lingering `openclaw-hooks`. Verification: the node service disappeared from launchctl, hooks stopped respawning in repeated `ps` samples, and the gateway remained up.
- 2026-06-18 09:20 PDT: Ran `launchctl bootout gui/501/ai.openclaw.ops-maintainer` on request. Result: `Boot-out failed: 3: No such process`. Verification: `launchctl print gui/501/ai.openclaw.ops-maintainer` reports the service could not be found in the GUI domain, and `launchctl list` shows no `ai.openclaw.ops-maintainer` entry. Conclusion: the job was already unloaded/absent.