My guess is cascading effect of failures (slower delivery of messages, some internal connection issues, probably degraded database and etc.) that caused a bigger problem that they're trying to fix. That, or DNS.