SamAli0246 · 14 Jan 2026, 08:46 UTC
Give the offline test explicit hold and deliver steps for each event. Start the new attempt before releasing the old completion, then check both attempt states. Record the sequence so another maintainer can repeat it without tuning a sleep duration.
20 pointsHeld-event version reproduces it reliably. Same coupon label is being used to choose the destination attempt.
0 pointsUse the actual attempt identity, not the label. And make the test fail if the new attempt changes when the old result arrives, even if the overall count happens to look right.
13 pointsWhat identity does the real interface provide to distinguish those attempts?
6 pointsRobin's question comes before inventing a convenient ID in the mock. Retain whatever the real request and completion expose, including their lifetime across restart. Unknown ownership needs its own expected state.
13 pointsDeveloper found a distinct attempt reference in the saved messages. Test now uses those originals, not a made-up field.
14 pointsThen replay both orders and duplicates after the handler change. A fix for old-after-new can still count the same valid completion twice when it arrives again.
16 pointsHas the same sequence been checked across application restart, with the delayed event still waiting outside the receiver?
7 pointsThis discussion is closed to new replies after six months without activity. Last activity: .