Aglet

How to learn from rate-limit scope and recovery rules

Rate-limit learning records what is counted, which actors share protection, how reset becomes visible, and what a safe retry means. Keep the critical workflow, fairness assumptions, response contract, and observed boundary together for future policy or traffic changes with clear operational ownership.

Keep the lesson for the next incident

  1. Document the limit contract

    Describe actor, credential, resource, route, request, concurrency, and time-window scope; allowed and rejected outcomes; reset evidence; response semantics; retry ownership; and duplicate prevention. Explain which critical actions require recovery. Keep examples structural and free of credentials.

  2. Keep a scope and reset journey

    Preserve checks for below-limit, boundary, over-limit, reset, separate actor, separate resource, noisy neighbor, critical action, and safe retry cases. Define expected statuses, messages, metadata, final state, and fairness. Assign an owner for interpreting policy changes and updating the journey.

  3. Recheck after traffic or policy changes

    Review representative cases after counter logic, identity scope, windows, route costs, concurrency, client retry, or response metadata changes. Compare protection, reset, fairness, and recovery again. Add a new case when a workflow has different durability or urgency rather than sharing one generic limit.

What to carry forward

The learning record should retain counting scope, reset meaning, response and retry ownership, fairness behavior, critical recovery, and recurrence triggers. It is complete when future engineers can distinguish intended protection from accidental denial and can verify safe recovery without duplicate effects.

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