mdnsd needs to try harder to send goodbye requests
Disabling mdns can result in ghost entries remaining in the caches of clients. This is because although SMF stops services in the correct order (dependents go down first, then mdns gets sent SIGTERM) mdnsd is quick to break out of its MainLoop. Because mdns is breaking out of its MainLoop the very instant it receives SIGTERM it does not have a chance to process any goodbye requests that might have come in from the dependent services as they exited. The solution in my view, is to have mdnsd have one last go and processing outstanding requests before exiting.