We have an erlang cluster that runs all our services @ http://thecadmus.com

High number of processes distributed across multiple nodes with supervisor trees taking care of them all has worked well for us.