or just cluster over several other providers and make sure to test failover

the main problem with azure/aws/gcp is that their parent companies will track everything you and others do, and then close every related thing because they have enough market share to not give a fuck about you.

There are more than enough blog posts of enterprise customers running into random issues with the big thing, buying "big three, promise edition" isn't much of an assurance.

I'd suspect most of those customers have rapid migration plans written up. For a reason.