AWS doesn’t have the power to wipe people out that don’t tie their stack to them. Parler could’ve avoided this, it’s not AWS fault that Parler could t move.
How AWS-specific your code needs to be depends on which AWS services you're using and how you use them. An EC2 instance is just a Linux system, as is a Lightsail system.
Other systems like MinIO and Ceph have S3 interfaces for object storage, so the code could be kept. Operationally you'd need to retrieve it from Amazon S3 and re-store it. RDS is an export, transfer, and import away from being loaded into a normal database server, but that can take quite a bit of time.
Getting geo-sensitive DNS and implicit testing of backends out of Route 53 isn't so simple. Getting stuff out of Lambda and onto a commodity platform isn't quick and easy. If you have data in Glacier there's a substantial lead time to recover your first byte.
Depends how much prep they did beforehand. Lots of businesses have contingency plans for negative situations including needing to leave AWS, and they're far less likely to get booted. I’m not sure why they didn’t just start with Epik; they already host Gab and won’t give a crap about what Parler does.
FWIW the Parler CEO claimed 12 hours, an obvious falsehood, and that’s now being used against him in his lawsuit against AWS.
Most probably couldn't. But most will likely never have to think of that. Parler's inherent value proposition increased this risk and they were foolish not to consider and plan for that.
Risk management is part of the job. Businesses that don't do it properly will often fail. Sometimes this is an unfortunate thing, sometimes it's a happy accident.
