I searched through both the capture agent logs and the core logs and not one error came up. Although all of the recordings ingested and processed correctly, each recording remained in the the ingest-tmp dir (/data/content/workspace/collection/ingest-temp). The good news is that random ingest failures haven't appeared.
Could the cleanup operation delete the associated zip in ingest tmp if it still exists?
Chris, I assume the same thing is occurring in 1.2.
I'm testing this on 1.1.x. Changes should already be on 1.2.