Prediction: the final postmortem will say "someone pushed a bad config", just like most of the previous postmortems (and most of the internal postmortems as well, for Borg-based services). This is the cause of most other outages in other cloud providers as well. A really hard to solve problem.
Multiple regions seem to be affected though. Wouldn't it make more sense to start out config pushes with a single region before expanding it to avoid these types of wide outages?
