The kuberesolver is faster because it receives events when a pod goes down? So in the event there was a stale spicedb server in HAproxy the service will timeout, or just not get a cache hit? We will just suffer from long requests?