Currently it is not possible to have a meaningful health check if your config sets protocol to TCP. I think it would be useful for many different services if you could provide a url for the healthchek rathern than just rely on the TCP connection. A fairly simple use case is having an https server consiting of a reverse proxy and an app backend, with the TLS connection terminating on the server, rather than the NB. In this scenario, the TCP connection check is useless, i.e., if the backend on a node goes down the NB will continue to route requests to that node, and you will get 502 or some such errors from the reverse proxy.