A week or two ago there was a conversation in #erlang on Freenode about wanting ChaosMonkey style testing to test the resiliency of BEAM applications. So I threw something together recently based on this library.
Once turned on, havoc will begin killing random processes as well as TCP and UDP connections in order to test that your application will properly handle failures. The only caveat is that it currently only works on the current node. I will be adding a feature shortly that will allow you to kill things on any node within your cluster.
Let me know if you have any questions!