Like Egate says, ubnt is not the solution. Even if their stuff might work easier in ´plug and play´ modus.
Frst of all, their hardware is not as reliable as routerboards. The failure rate is much higher.
Second, it has only some 60% of the options as a fully squeezable ROS like MT has to offer.
Third, I do have two very short range backhauls with loco5’s (airmax) and they gave me same disconnects as what we see in MT’s NV2.
The links are only stable by using channels way out of range of what normally is in use… 
and disabling the airmax system.
But, like Chadd says, their are more problems. This ´ghost´ unit is not the only one.
I am busy making a inventory of all my units to see which units still give what kind of problems.
My whole network, +/- 250 units is now v.5.4 with updated firmwares.
A rough estimate is that I have almost 10% (!) units still having the port flapping issue. Although I reported it a while ago it was gone, it is just back. Only not that often any more as before and on much lower amount of units. It took me some time to realize it is still happening.
And about 5% is still showing disconnects with all kinds or reasons. I know this has all to do with the many channels in use in my area and thus interferences, but the strange thing is that some units give the problem while neighbours with the same unit and running of the same AP only 5 mtrs beside of it (seen from the AP, so this particular unit is in the middle) have not a single disconnect!
About the port flapping issue still around I send MT a support request and they came back with a .pdf file on how to check several physical statuses of the ports and other parts of the routerboard.
Like suddenly some 30 units under 5.4 now have physical problems where they haven’t had these half a year ago! (in some 4.x versions..)
And it is not a hardware related issue. It happens on rb411, rb433’s, SXT’s, rb133C3’s, rb122’s and a rb532A.
The problem is that 5.x family and NV2 does improve management and help me stabilize my network much more in regard of other issues I had in normal 802.11 but it comes with new problems… which is indeed very disappointing!
I have to find some more time to really analyse (wearing badge reading “unsolicited MT field test manager”) what is going wrong on the several units and under what conditions it happens and when etc. Some units obviously show in the register and logs to be disconnected but the clients never notice anything. When asked they make regular skype phone calls and never have a single problem. They tell me about how good it is! (So, althoug it sounds nice, it only confuses me more, the logs really tell me it should suck!)
I have one unit disconnecting almost 200 times a day! Client doesn’t notice anything and makes regular skype calls they say… beats me… 
It’s been quiet for a while on the new release of rosv5.x so I only hope MT guys are really trying to nail the issues and wait with a new version until they really have solved some of the issues still around…
In the meantime we just struggle on and I can only find grieve in the fact I hardly get complaints from client. And not because they are tired of it. … Last Tuesday I had a failing rb333 (in a backhaul link) that over 1 hours started to disconnect for whatever reason and finally collapsed beyond recovery (yes, completely DEAD) so I had to replace board. But within half an hour we receive 50 client calls about the failing internet. So clients DO know to find me if they have no service.
