We hit a silent cron outage where no jobs finished for ~200 minutes, even though openclaw cron status --json stayed enabled and nextWakeAtMs kept advancing.

Symptom

• Fleet-wide largest finished-event gap (last 24h in our run logs):
  • from 2026-02-06T04:49:38.544Z
  • to 2026-02-06T08:10:31.957Z
  • gap 200.89 minutes
• During the gap, scheduler remained enabled and wake timestamps advanced, but run logs were not appended.

Suspected root cause

In timer wake path (onTimer), the cron store is force-reloaded before due-job evaluation:

• onTimer calls ensureLoaded(state, { forceReload: true })
• ensureLoaded calls recomputeNextRuns(state)

That recomputes nextRunAtMs from current time, so jobs that are due at wake can be pushed forward before runDueJobs executes. Repeating this on each wake can defer jobs indefinitely without a hard scheduler error.

One-line fix

In onTimer, avoid force reload:

• from: await ensureLoaded(state, { forceReload: true });
• to: await ensureLoaded(state);

Local confirmation

Applying that change locally restored cron execution immediately in our environment (new finished events resumed for 10m/15m jobs right after restart).

Environment

• OpenClaw runtime observed: 2026.2.3-1 (d84eb46)
• Install path: Homebrew global install

Happy to provide additional logs or help test a proper upstream patch if useful.