That's still not answering many questions. I hope they publish a full analysis at some point.
Since they're using default provisioned EC2 instances, it's likely that the developers don't necessarily even fully understand their performance degradation. They just expect the service that they pay for to work properly.
It's true, but it's not what I meant. They wrote "after a host was patched". This is ambiguous. So they mean the host as in instance, or host as in AWS host machine? Did they just reboot it get on the new/updated VM host, or did they rebuild to include the PTI fixes as well. Did they upgrade anything, or did everything else stay on the same version.
From what I can tell, Amazon isn’t giving people deep technical information. They’ll just send you an email telling you which instances are being forced to restart and when.