Check whether the outcome improved
Test freshness cases
Use bounded fixtures for a normal change, a delayed delivery, an out-of-order version, and a source with no new update. Assert expected destination state and freshness classification for each. Include equal timestamps with different versions when version ordering matters.
Measure end-to-end lag
Compare source change time, delivery time, processing completion, persistence time, and read time against the documented interval. Review the oldest and newest affected entities, not only an average. Explain any lag that falls outside the intended cohort or recovery window.
Check recovery boundaries
Run the bounded replay or correction path and verify changed values, versions, counts, and downstream views. Confirm repeated execution does not create extra rows or regress newer state. Retain evidence for entities that remain stale so unresolved work stays visible.
What to carry forward
Accept recovery when known source changes reach the intended destination within the stated window, out-of-order inputs follow the version rule, quiet sources are classified correctly, and remaining lag is explained. Keep verification open for unobserved stages or broader source cohorts.
Keep the decision with the work.
Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.
Create an account See the product workflow