This could be a one off but this was a bit of a disaster for me, pushed it to our MR46s running 28.3 and straight away all of them started a weird cycle of being offline/uncontactable for 2 hours in Dashboard, then online for 2 hours, off for 2 hours, etc...
Dashboard was showing an error for every AP talking about them being plugged in to a different device for LLDP and from the switch level (Cisco Cataylst) I was seeing down/up flaps on the MR interfaces in line with the 2 hour cycles + duplicate CDP neighbour entries, one would show Port ID 0 and the other Port ID 1.
Had to roll back to 28.3 and they're stable again, won't get a chance to get a unit in the lab until Tuesday to recreate and get in touch with support.