It wasn't actually down from overuse, the node.js server hit an unhandled error and fell over. And because this is the first time I was seeing real usage, I didn't have a supervisor in place to restart it automatically. That's handled now, of course.