I've had this happen to me multiple times. Luckily I use AWS for batch research jobs, so I'm not losing any money when things break. But when researchers come

I've had this happen to me multiple times. Luckily I use AWS for batch research jobs, so I'm not losing any money when things break. But when researchers come to me with AWS problems even though I've not changed anything, I usually spend 10 minutes on one of the nodes for some cursory investigation and if it's not obvious what the problem is I tell them to wait a day or two before further investigation. 95% of the time, the problem clears up on its own. I've learned the hard way that spending half a day tracing SQS messages, dynamo tables, spot bidding, etc is usually a waste of time. That the AWS console flat out lies like this wastes so much of everyone's time. It's not even hard to fix, AWS could run internal test cases on every subnet group.