Power failure on the night of 1st killed two nodes. The rest died on the morning of 2nd. Restarted jobs but started wondering whether 'persisting until prevailing' is a viable option under the circumstances.