Agreed. We run an Akka cluster with 5-6 nodes as our backend, clocking in at 15Kloc. It's been running for months now with no problems. If we need to add cap

Agreed. We run an Akka cluster with 5-6 nodes as our backend, clocking in at 15Kloc. It's been running for months now with no problems. If we need to add capacity, we just provision a new EC2 instance with Vagrant/Chef and it auto-joins the cluster. Once we're done, we can remove it gracefully and the rest of the cluster just deals with it.