Comment by arafsheikh

Typical case of degraded mode of operation, I presume. That’s how well engineered systems should behave when they’re about to go down.

My guess is that this particular outage was caused due to a scaling constraint somewhere in the stack.