An asynchronous instance failed after a temporary outage, and the target is now healthy. Which approach is appropriate?

1. Discard the message and expect a subsequent replay from the same error record.
2. Submit repeated replays while a prior resubmission remains in progress.
3. Assume Oracle Integration automatically resubmits every failed message.
4. Verify recoverability, assess duplicate side effects, and use supported resubmission.