In the event that an upload is completed before any bytes are actually written to the stream, StreamTransferManager instead calls putObject to create the empty object. That's fine, except the multipart upload that was initiated with the call to getMultiPartOutputStreams is never completed or aborted and thus might persist in the bucket forever.

This can be worked around if the caller calls abort directly, or mitigated if the bucket has a lifecycle rule that aborts incomplete multipart uploads, but it would be more appropriate for the library to abort on its own since it knows it won't ever write anything more to that specific upload ID.