Cron 防僵死自愈 + in-flight 任务重入队

• You can upgrade OpenClaw to a build containing PR #14430 and restart gateway.
• You have permission to create/edit cron jobs and inspect cron status output.

1. Upgrade to the build with PR #14430, then restart gateway to load anti-zombie scheduler logic.
2. Create one recurring job and one one-shot reminder (`at`) as canary probes.
3. Simulate timer disruption (e.g., staging event-loop stall) and observe `cron list` for overdue timestamps.
4. Wait for anti-zombie window (~60s) and verify scheduler re-initializes without full process restart.
5. Confirm stale in-flight jobs are cleared/re-enqueued and one-shot reminders are retried instead of dropped.

• PR was in review at collection time; final behavior/flags may change before merge（需验证）.
• Do disruption tests in staging first; forced event-loop stalls in production can impact unrelated workloads.

This tip is aggregated from community/public sources and preserved with attribution.