Since you ruled out gzip itself as the culprit, this looks to me like a pipeline problem, and not exactly a gzip failure - which means the trouble would be in ArchiveProcessor.
In other words ArchiveProcessor may not be handling the pipeline correctly (a bug! ... perhaps). There are at least 2 other possibly related issues to be found on this site:
http://answers.splunk.com/answers/57272/large-data-archives-zip-being-corrupted-on-indexing.html
http://answers.splunk.com/answers/132045/error-archiveprocessor-with-zip-files.html
these refer to zip files, but i wonder if it might be common cause, specifically pipeline handling in the ArchiveProcessor logic.
As an aside, i'm completely unsure if your issue is related, but these are food for thought:
https://blog.nelhage.com/2010/02/a-very-subtle-bug/
http://bugs.python.org/issue1652
... View more