Same answer, wrap you usage and use the wrapper for agonistic testing & coding.

That’s the entire point: my domain code ‘stores’ something, my implementation decides if that is S3 or a fixed kind of error or writes to a disc.

If my domain uses a third party app that needs an S3 API, and we see not pointing it at S3, then I am writing an s3 API to test that dependency and ensure the third party code works regardless. Normally I’d call that “their problem” and call their service with the same wrapper as above.