Hoo boy.

There are regions and then there are regions. Imma thread this one. https://twitter.com/daveadams/status/1285987768749961220

us-east-1 is the original region in many respects. It's comprised of many dozens of buildings, with largely separate power, network, cooling, etc. infrastructures.

It's multi-homed to multiple providers.

It's MASSIVE.

(This is, incidentally, why the somewhat-maligned http://gaslighting.me  status page gets so much flak. Something is ALWAYS broken in something that large.)

So what does "us-east-1 becomes unavailable for a protracted period of time" look like?

We're talking about ~100 square miles or so of Northern Virginia being swallowed up by the earth, effectively. It's hard to imagine such an event in isolation.

Some things are single-tracked through us-east-1. I mean, you don't get independent AWS bills sent from each region!

Capacity in other regions would be tight--it is any time there's a failover event at large scale. This is partially why having capacity for DR pre-provisioned is so important. The control planes get... congested.

Some things would absolutely break, yes. But most planning accounts for regional outages lasting hours, not months.

Some disasters are orders of magnitude likelier (do not misread "likelier" as "likely):
* @awscloud boots you off of the platform
* Your credit card is declined and you don't get the notifications
* Amazon decides it's going to Google its AWS division

Some companies need to think about these things. Most do not.

If you're CapitalOne, you still need the financial system to work if AWS sees massive disruption.

Twitter For Pets can probably weather the storm.

One group that's extensively modeled these "AWS region becomes permanently unavailable" scenarios?

Investment firms.

Their entire business is determining risk and hedging against those risks.