If the queue is SQS, retries due to job timeouts are different from retries due to errors.

Timeout retries are retries due to visibility timeouts.
Error retries are retries due to messages being re-stacked in the queue.
I think the default timeout retry behavior should be the same as the error retry behavior, but I would like to know if it is your understanding that the control of timeout behavior is left to the job implementer.