Since the 29.11 incident when most of Ukko cluster's nodes got hung, half of the cluster nodes have now been rebooted to a new kernel and other half are still running old kernel. This will be fixed by rebooting all nodes still requiring restart with a schedule described here. So when you are starting new jobs, please consider using those nodes which have already been restarted. You can see the list of nodes and their status at the hpc-report page.