Failures are alerted rather than silent, and every payload is logged so the affected records can be identified and replayed. Under a support agreement there is a defined response time; without one, the alerting and the logs still allow your own team to act.