Comment 2 for bug 1783315

I can't be absolutely sure when this regressed, since such soft timeouts aren't very common; there've only been three since the last rollout. One thing I noticed was that a soft timeout on a PlainPackageCopyJob on 2018-07-16 didn't cause celery to hang, but soft timeouts on GitRefScanJob instances did. My current suspicion is that it has something to do with urlfetch's thread handling.