The job retry counters don't seem to be working, at least for ParsoidCache* jobs:
gwicke@tin:~$ mwscript showJobs.php --wiki=enwiki ---type=ParsoidCacheUpdateJobOnDependencyChange --list | grep ParsoidCache | grep -c '=201401'
gwicke@fluorine:/a/mw-log$ grep ParsoidCache runJobs.log | grep -c '=201401'
Normally I'd expect those jobs to be retried three times only (default setting for 'attempts' in puppet/modules/mediawiki/templates/jobrunner/jobrunner.conf.erb ). Instead, these jobs have been around since January. The 86 jobs have been executed 4006 times today alone, which works out to more than three retries per job.
Example runJobs.log line:
2014-10-08 08:35:49 mw1007 enwiki: ParsoidCacheUpdateJobOnDependencyChange Module:Message_box/configuration pages=array(12) rootJobSignature=dd78b0f638bd2f608d82c1e31574b46271060622 rootJobTimestamp=20140127010733 table=templatelinks type=OnDependencyChange t=14415 good
t=14415 would be over the default timeout of 3600.