Testing the DataDrivenAg instance, simply clicking a single endpoint link (cloud9, in this case) resulted in 800+ check_token calls in a 30 second period. Since I'm using check_token as a method to log last activity, this is causing the apiserver to hammer etcd with updates. I've disabled this for now, but it would be good to diagnose why this is happening, as it could be the cause of instability.