Healthcheck lock contention: manual rerun during active launchd run refreshes nothing
brain-distill · confidence medium · status active
tags: infra healthcheck launchd locking ops
The official healthcheck script (/Users/dwayne/bin/codex-hermes-openclaw-healthcheck.sh) serialises on a lock. Manual invocations on 2026-08-28 and 2026-08-29 ran while the launchd-scheduled instance held the lock (pid 56279, pid 77588). Each manual run exited rc=0 but refreshed no artifacts, so operators kept reading stale dashboard and contract-audit state and re-investigated healthy services. Rule: before a manual rerun, check for an active healthcheck process (ps for the script or its launchd label); if one is running, wait for it to finish and read its artifacts. Treat ran-script-rc0-with-no-change as lock contention, not success.
Sources:
/Users/dwayne/clawd/memory/health-remediation.log