Doing it wrong? If one DC going down, takes out your "cloudy" application, you suck at making cloudy apps. Surely? It's a pretty big SPOF if you are trying to be highly available.
If a single region is a single point of failure, isn't that "doing it wrong"?