Stack Overflow proves we're all doing it wrong. They probably serve close to half a billion requests per day using a server architecture that's very close to what this old man grew up on. This article slightly clarifies their use of static calls, but I'm curious how they balance this with testability going forward.
The 3rd week in a row reporting an AWS outage. This one again occurs in US-EAST-1. If you lost data in this outage, it's because you decided it wasn't important enough to set it up right.