Comment by nieve

I think this is actually the desirable behavior. If you've got very long-running connections and you want to force a switch-over you can always drop the connection even if you don't have some in-band mechanism. If you're constantly watching for a DNS change all it takes is one DNS failure to kill a connection (or all connections). In general transient network issues are probably going to be more common than IP changes on your infrastructure and issues caused by the former are harder to debug.