GS108Tv3 keeps downing port 8

fw 7.0.8.4 (current at time of writing)

I am using GS108Tv3 switches purely in order to implement a VLAN so that I can segregate traffic over an existing ethernet radio link.

the radio link has been in use as a way simply to extend a UK consumer broadband connection into a rural area. I wish to now use that same radio link to carry a separate broadband internet connection in a similar fashion into the rural area. basically 2 different subscribers paying for 2 different broadband services in their office, but wishing to use said office broadband service at their homes a few miles away in remote areas. Clearly we want to avoid the situation where we have 2 DHCP servers conflicting with each other.

A way to do that is to put the new subscriber on a VLAN in order to keep his traffic separated from the existing subscriber. we will (for simplicity) keep the existing user un-VLANned. (or, rather, on the default VLAN)

I have established that the radio infrastructure carries VLAN data (that is, it does not strip out VLAN data on packets)

it may be possible to do the VLAN-ning using only the radio equipment (ubiquiti) but the VLAN functionality on the radios is somewhat basic so I have chosed (with guidance) not to do this.

instead, I have chosen to use GS108Tv3 "managed" switches (see topology pic attached) in order to implement the VLAN. 

in short, traffic entering the switches (on port2) is tagged (using PVID=>10) into VLAN 10, and traffic exiting the switches (on port2) is untagged. This makes port 2 our "access" port, at either end. in other words,

Port 2 is a member of VLAN 10.

Port 8 is a member of VLAN 10, but there are no PVID assignments on Port 8.

port 8 is the radiolink port. by being a member of VLAN 10, it will accept any VLAN10 traffic (which in this case will only be coming from its partner switch at the other end of the radiolink). Likewise, it will pass the VLAN 10 traffic from port 2.

I cannot actually rembr now, I dont have the devices in front of me, but I *think* port 8 is a member of ONLY VLAN10, and this is to stop any other traffic from the radio system mutually interfering with user equipment.

There may well be refinements in the VLAN scheme I am attempting to use, whatever, it seems to basically, work: there is no interference to either of the users sharing the radio infrastructure, and each user is able to use their respective broadband service at their remote locations etc.

the problem is that frequently, and irregularly, port 8 on the GS108Tv3 switches just fails, with associated loss of link,

Powercycling the switch always fixes the failure (and restores the link), but the same thing happens, between 1 hour and 2 days later. there seems to be no specific time interval.

I have logs running on the ubiquiti radio equipment and they are rock solid.

I don't entirely understand what I am looking at in the various logs within the GS108Tv3 switches, but I attach a recent log capture which does not appear to be very helpful:

8 hours after power-on, "Interface GigabitEthernet8 link down"....

10 milliseconds after that, I get something more interesting; "loopback-detection error detected on GigEth8, putting GigEth8 in err-disable state"

initial googling of loopback detection error shows me that this is a "thing", but I dont understand what. I also note that there is some facility within the switch where I can DISable L2 error detection. I have not tried this yet but it smells like something I ought to do.

What could be causing port 8 to suddenly go down? I was hoping for some log entries immediately prior to this down entry which could give me a clue. 

is it possible that the loopback-detection error is actually causing the port to go down? the log reads more as if the port mysteriously goes down, causes the loopback detection error and then this loopback error ensures the port is not automatically retried.....