I'm not sure I understand the issue here: If you have an at-least-once queue, it might still be reasonable for the application to directly write to a transaction log somewhere that can be reconciled for auditing, but that audit should always pass (even under fault conditions): there is either a design flaw in the system or malicious insider tampering if the queue does not deliver the message. Is it necessary to audit for that sort of thing on a sub-day level?
It doesn't seem like an appropriate place to put a wait-and-retry loop -- it absolutely breaks the entire point of queueing if the sender does not consider the task completed and flush all of its state once the message has been successfully enqueued.
It doesn't seem like an appropriate place to put a wait-and-retry loop -- it absolutely breaks the entire point of queueing if the sender does not consider the task completed and flush all of its state once the message has been successfully enqueued.