1.) Start up a vanilla 3 node set on say ports 2400, 2401, 2402.
2.) $ mongostat --port 2400 --discover 5
3.) All nodes are discovered and report stats - everything's happy
4.) Shutdown node on port 2401.
5.) mongostat continues, simply reports "no reachable servers" for the down node - still fine.
6.) now with the 2401 member still down, stop mongostat and restart with same command as step #2.
So, if any node is down when first connecting, the whole command fails instead of continuing on with the 2 good nodes like in step #5, which is what I would expect.