Not necessarily a single point of failure. There could be standby servers that were promoted to primary when the primary failed, then fell over with the huge spike in load. Or they were load balanced, multiple failed, and then the others tipped over.

In those cases, there were mitigations, but the failures cascaded.