Some running jobs are being interrupted and returned to the queue more often than we consider acceptable. Interrupted jobs retry automatically - you do not need to resubmit - but affected jobs will take longer to finish than usual.