A working reference / no signup
Replay Readiness Checklist
Before resetting progress, establish what can repeat. Use these questions to document one flow, its evidence, and who can stop it.
This is a design and runbook aid, not authorization to replay, an executable procedure, or a certification of safety. Unknown answers require investigation or escalation, not a checked box.
Checkboxes are optional local markers. They are not saved, submitted, or tracked. Keep your answers and operational data in your own approved system.
Name the boundary and the owner
Prove which intent can repeat
Locate progress relative to effects
Reconcile unknown outcomes
Bound the replay workload
Define stop conditions and evidence
Practice before applying it
Try the Async Failure Lab to see why a lost result can produce a duplicate effect. Its synthetic trace is an illustration, not evidence about your system.
The underlying contracts are discussed in AWS on idempotent API retries and Stripe on idempotent requests. Neither a checklist nor an operation key creates an atomic transaction across independent systems.