Check whether the outcome improved
Replay matched fast and slow shapes
Use identical routes, methods, payload structure, result counts, account classes, cache states, and dependency conditions from the baseline. Record total duration and each available interval for repeated runs. Compare the spread and the central behavior, not only one improved slow request.
Check contract and status behavior
Compare response fields, ordering, pagination or continuation, status codes, error objects, and headers relevant to the client contract. Exercise validation, authorization, empty, and dependency-failure cases. Confirm the change does not turn a slow but valid response into an incomplete or misleading success.
Exercise retry and waiting workflow
Replay the action that waits on the API, including a timeout or delayed response where safe. Confirm retry controls have clear ownership, repeated submission does not duplicate state, and stale responses cannot overwrite newer work. Record client feedback from request start through final usable state.
What to carry forward
Classify the result as more predictable, regressed, unchanged, or inconclusive with distributions, interval evidence, contract checks, and workflow behavior attached. It passes only when representative request shapes show reduced harmful variance and clients still handle success, delay, and failure correctly.
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