Huh? i have tons of NLB, everything we just issue drain stop request then perform upgrade/maintenace on the node, we have never experienced any issue with regards to the nlb cluster. When you drain stop it, all current session will stay valid until session end, while new request are never route to the node. You are not seeing this?
And of coz NLB is network layer protocol, it doesn't have visibility to your application stack, so if IIS is down and NLB is still up, request will still route to the box but with error as IIS is not fulfilling request. You can always get some 3rd party tool or write a script to query IIS status if it is down, then issue NLB stop command to pull the node out of the cluster.