Check whether the outcome improved
Define response policy
Write expected behavior for a rate response with reset timing, retry-after, no timing header, a normal success, and a permanent rejection. State max attempts, jitter or spacing, queue state, and visible outcome for reads and writes.
Run single and burst cases
Execute the policy with one caller and a bounded concurrent group. Compare attempt timestamps, concurrency, response headers, and queue state. Confirm callers do not synchronize their next attempt and that one work item retains one identity through recovery.
Check recovery boundaries
Simulate a reset or successful response, a continuing rate response, a cancellation, and a process restart. Verify optional work can pause, essential work remains visible, and a missing timing signal does not produce an unbounded retry loop.
What to carry forward
Accept when backoff honors available evidence, attempts remain finite, concurrent callers avoid a retry wave, and work identity survives recovery. Keep remote quota behavior partial when it is not controlled. Record policy, fixtures, response headers, and untested caller classes.
Technical background: GitHub documentation.
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