Decide where the work belongs
Rank the action exposed to the tail
Name the user action that waits for the response and distinguish a delayed background refresh from a blocked save or decision. Record whether people retry, navigate away, or submit again. Unpredictable waits that invite duplicate actions deserve more weight than equally long invisible work.
Group reach by request shape
Compare route, method, payload fields, result count, status, account size, cache state, and release window. Use distributions and request volume together; a rare extreme and a common moderate tail imply different queue decisions. Mark dimensions with insufficient observations as uncertain.
Choose a bounded variance slice
Select immediate investigation, a route-specific reduction, or collection of one missing interval or request group. State the evidence that changes priority, such as a new release boundary or slow traces sharing a data shape. Avoid setting a universal target before the affected action is understood.
What to carry forward
Produce a queue decision tied to blocked work, tail reach, request volume, and evidence quality. Include retry or duplicate-action risk when observed. The item is ready when the next engineer has a representative fast and slow request and a reason the variance merits capacity.
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