We found the source of this and were able to solve it. All instances are healthy now.
We had moved some instances last night and there were orphaned records that caused connection issues.