What sort of files are they? Any clues from the filenames? Is it possible to use the file command to see what sort of data they contain? Failing that, what sort of operations are you doing in your pipelines? I can see in the source that Data Collector will create files in /tmp when working with whole file data format, but it should be cleaning up after itself. If you can pin down enough information for us to reproduce the problem, please go ahead and create an issue.