We were experiencing the same exact mDNS bugs as you - rebooting a random controller almost once a day for a few weeks... occurred primary when doing a "sh switchinfo" or "wr mem" on the controllers or the master who then sent it to the locals and caused them to go in their funky state.
We upgraded to 6.3.0.2 last night and things are working well - I can finally do a "wr mem" and not plan to go reboot the thing :)
Been testing all morning random stuff that beore would cause the controllers to break and everything seems solid.
I recommend it. We've been prepping for 6.3.1 for quite some time but that date has been changing for quite some time so to be continued...