Keep the lesson for the next incident
Document limit handling
Record response classes, usable timing headers, backoff calculation, maximum attempts, jitter, work priority, and unknown-header fallback. Define separate treatment for reads, writes, and reconciliation. State whether concurrency or remote quota is the owner of each limit.
Keep burst fixtures
Retain single, parallel, reset-guided, missing-header, continuing-failure, and restart cases. Store expected attempt timing and work identity without credentials or payloads. Add the local fan-out fixture if it was the source of amplification.
Review pressure signals
Watch concurrency, rate responses, retry counts, queue age, and synchronized attempts by client revision. Assign an owner and threshold for a bounded investigation. Close the follow-up only when the policy still protects essential work after caller or transport changes.
What to carry forward
Close learning with limit classifications, burst fixtures, retry and capacity owners, and recurrence signals. Keep external quota and secondary-protection limits explicit. The useful outcome is a measured finite response to pressure, not a promise that rate responses will disappear.
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