Thanks Jake! We'll definitely write some posts in the coming months about our architecture. We never intended to be running a high-availability service, and we have learned a lot of lessons along the way.