Allow S3 Consumer to be configured to not return/delete "empty files" which S3 uses as a way to simulate a folder structure.
Since S3 object Content-Type meta-data tells whether the object is a "directory", this can be used to identify and skip such objects. It does mean that the number of objects returned could be less than the maxMessagesPerPoll. As long as this is not the default and the docs explain what to expect, I believe this would be okay.
Wanted to create a JIRA to see if there was an architectural objection to adding such a feature. If not, Ill try submitting a PR.