The handover test people forget: replay the same event

A successful first run is only one part of an automation handover. Ask what happens when the same event arrives again. A duplicate webhook, manual retry or restarted job can turn a routine recovery into a second customer record or repeated action.

Make that question testable. Choose a synthetic event with a stable identifier such as lead_id. Record the first expected result. Replay the same identifier, and record the second expected result separately. “The workflow works” is not an acceptance condition; “one CRM record exists and the second event is marked duplicate” is.

Name the person who can verify the result and the evidence they should inspect. Then document the boundary: which failed events can be retried, which need investigation, and who owns that investigation. A blanket instruction to retry everything is not a recovery procedure.

For the handover, write four lines: input, expected result, reviewer and evidence reference. Add one more: the next action if the result differs. Use synthetic data for the exercise. This does not prove every failure mode is covered, but it makes one important assumption visible before the builder leaves.