Apparently this job doesn't have a timeout, so some jobs were running for 8 hours until jenkins had an internal error and killed it. These should have a timeout of no longer than 30 minutes honestly.
$ grep -L timeoutMinutes * analytics-refinery-release analytics-refinery-update-jars apps-android-wikipedia-periodic-test mediawiki-core-code-coverage mediawiki-phpunit-coverage-patch mwext-phpunit-coverage-patch mwext-phpunit-coverage-publish