Well, er, good, fortunately my point was always in favour of spanning regions, not zones, so I refer you back to my opening note about not conflating disaster recovery (i.e. "we lost the DC, now what?") with high availability ("something within the DC is down, shouldn't affect the application").

Because in that analysis, the AZs never have represented an isolated unit of availability. They are clearly locally interdependent and/or sharing infrastructure. Heck, not sharing a continental plate is pretty much my #1 criteria for a DR replica.

NB: choosing to deploy in us-east just says to me "I want to save bucks, I don't care that it's got by far the worst availability track record".