Audience question: when a POST depends on some entity that may have been updated between the initial attempt and the retry, should the retry send the new parameters, or stick with the original payload? Answer: apply optimistic locking — attach version 0 to the initial create; on retry the receiver compares the current version and either acknowledges (the first attempt created it) or rejects with a conflict. The sender can then decide whether to discard or continue its workflow (possibly firing side effects/events it still owes) without crashing.