openclaw - 💡(How to fix) Fix Cron: missed-run catchup/backfill on gateway startup [1 participants]

Problem

When the gateway restarts and a cron job's scheduled time has passed, the behavior is inconsistent:

• Sometimes the job is backfilled (if the gateway starts right after the scheduled time)
• Sometimes it is skipped entirely (if the previous instance already "claimed" the run but died before completing it)

There is no reliable missed-run catchup mechanism. If a job was scheduled during a gateway outage or restart, it may be silently skipped with no record.

Observed Behavior

Two jobs scheduled close together on the same day:

• Job A (18:45 Sunday): Gateway restarted at 18:45:32 → backfilled at ~18:50 ✅
• Job B (19:15 Sunday): Previous gateway instance fired it at 19:15:00, was killed at 19:15:35. New instance started at 19:15:38, saw 19:15 had passed, scheduled next run for 7 days later ❌

Job B was silently lost. No run record exists. openclaw cron runs shows 0 entries.

Expected Behavior

On gateway startup, the cron scheduler should:

1. Scan all enabled jobs
2. For each job, check if nextRunAtMs is in the past
3. If a job's window was missed (within a configurable grace period, e.g., 1 hour), backfill it
4. If a started record exists without a completed record (see related issue), retry it

This should be opt-in or have a configurable grace window to avoid firing very stale jobs.

Suggested Implementation

On startup:
  for each job where nextRunAtMs < now:
    if (now - nextRunAtMs) < gracePeriodMs:
      queue job for immediate execution
      log: "Backfilling missed cron: {job.name}"
    else:
      log: "Skipping stale cron: {job.name} (missed by {delta})"
      advance nextRunAtMs to next occurrence

Configuration

{
  "cron": {
    "backfillGraceMs": 3600000,
    "backfillOnStartup": true
  }
}

Environment

• OpenClaw v2026.3.11
• macOS (arm64)

Related

• #59301 (gateway restart loop should pause/defer cron jobs)