FWIW I initially went into ELB assuming it would solve a lot of my redundancy problems. And while it has helped a lot (I spread my frontend across 3 zones), I've suffered through a number of ELB failures or disruptions, including this latest one, which is one of the worst. Even with fully functioning servers that I can connect to individually, ELB was intermittently rejecting connections and failed to reregister instances. There's no silver bullet! Just prepare for failure and attempt to handle it gracefully, learning from each one. I suppose you should also think hard before you launch into a greater AWS budget to increase availability. Most of us are tempted to do that after each major incident--which is why Amazon can walk away from these events in a better position than before (until they have a genuine competitor that is).