Fictional example · Approved demo revision 1 · Testing failure paths
My API request timed out. How can I tell whether retrying will do the work twice?
Separate the result of the operation from the response you received. A timeout can leave the result unknown. For a small exercise, write down what the client can observe before and after a commit, then give retries a stable operation identity. Check which boundary actually records that identity.
Resources to review
Fictional example · Approved demo revision 1 · Testing failure paths
How could I test a worker that receives the same event more than once?
Start with three cases: one delivery, a repeated delivery, and a committed operation whose acknowledgement is lost. Compare the resulting records, not only the number of calls. Add a conflicting payload using the same identity, and decide what the worker should reject. State the scope and retention assumptions of the guard.
Resources to review