Missed Job Handling

What happens when the scheduler is down for an hour and 4 different jobs should have fired during that window?

Two strategies:

Skip (the default for cron): only fire the most recent missed instance, or none at all. Avoids cascade overload after recovery.

Catchup: fire each missed instance in sequence. Used when each instance must run (billing aggregation, log rotation). Bound the catchup window so a year of downtime doesn't trigger a year of jobs.

Kubernetes CronJob exposes this via startingDeadlineSeconds: if a job is missed by more than this duration, skip it. AWS EventBridge does the same.

Trade-offs:

• Skip is safe but loses work
• Catchup preserves work but can spike load
• Catchup with limit balances both — fire up to N missed instances