A manifest worker can finish its update and then fail before it sees a deletion response. The next delivery may represent work already completed. Generating a new business operation for that attempt would turn a recoverable transport uncertainty into a repeated side effect.
Design the consumer around stable intent, durable outcomes and a deliberate retry policy. A temporary dependency failure differs from invalid input that will fail the same way forever. These tasks combine real local receipt transitions with design decisions about repeated work. They do not execute a distributed consumer, expire visibility periods or route messages to a live dead-letter queue. The goal is to state the contract an actual worker must satisfy and the evidence an operator needs when that contract is uncertain.
Settlement and external business effects have separate failure paths.