Hi Sarah,
Did you happen to try logging into one of the non-master nodes during the failure? I'm curious what they saw their role as during this?
Also, did you check 'show slot'? I'd like to know what the status of the non-master nodes was.
'Show log' from both the master and one of the failed nodes may be helpful as well, but since it was rebooted and the issue was a few days ago, there's a possibility we may have lost the logs during the failure.