Our network has a number of systems that are dynamically allocated IPs (which happen to be in the public range). These systems also have a local hostname, which is making it into the DNS resolver cache with whatever IP the system currently has. When the system gets a new IP, clients trying to connect via hostname are having to manually flush the DNS resolver cache before they can find the host at its new IP.
I believe that simply preventing the cache from including hosts within a certain IP range should resolve the issue. Working the other way, if it's possible to exclude non-fully-qualified hostnames that should also resolve the issue.
I've done quite a bit of Googling, but nothing useful is showing up.