My guess?
My guess? Rolled out newer DB software. It corrupted something. On rolling out older DB software, DB still corrupted so it didn't come back up properly. Restored DB from backup and (possibly) brought backup up to date using transaction logs.
Obviously for reliability for a "cloud" DB, it should have been possible to upgrade a small portion of the DB machines (perhaps just 1 machine at first), and see it fail, without really impacting service since the rest would be running the old DB software.
Biting the hand that feeds IT
