After a power failure hard rebooted the switch. Apparently all nodes came up without detectable network problems (?). Try starting a job on n0001 to see if it is stable now : No, still fails immediately.