Alright, so I narrowed it down. I am running node under the new cluster module. What is happening is one of the node processes is dying and it that is what is closing down the connection to the client. It happens right before the node process response.
This leads me to another couple of questions. I am running node under the Webstorm debugger and I can't get all of the cluster processes to debug. In fact, when I run node in a clusterd configuration as debug my client can't even connect to the debug clustered node server.
Is there a way to get debugging to work with a clustered node server in webstorm?
Ok, now for figuring out what is crashing... Is there a way in node to report out the crash cause and stack when a cluster child crashes?
Also, when you are running node in webstorm, how do you get it to write console output to a file?