We had minor storage issues earlier that caused one storage node to take over for another. This is usually a successful operation that you wouldn't notice, but this time some client compute nodes did not reconnect to the storage for some reason. Some jobs may have died and we are working to figure out why that happened.
