Our offline FR5 result tracker lets an old attempt's completion finish the newer attempt when it arrives late; how should I make that sequence repeatable so the developer can test the fix, rather than hope a random delay reproduces it?
Use a controlled event queue and test clock so you choose the order explicitly: start the first attempt, start the second, then deliver the first completion. Do both request and completion carry an attempt identity the tracker can compare?
They only carried the reused job label; the revised fake interface now carries a separate attempt identity, and the queued late completion leaves the newer attempt waiting. The real interface change still needs agreement with its owner.