Reliability and operability

Key question

Can the workload recover predictably without duplicating or losing work?

The first case will use AWS Step Functions to demonstrate bounded retries, idempotency, replay, and explicit failure paths.

Planned evidence

• Retry and backoff behaviour
• Idempotency validation
• Failed execution replay
• Actionable operational alarms
• Exam guidance compared with production guidance