"Wireless wire" (a pair of RBwAPG-60ad) periodicall stop forwarding

Greetings,

I have a Wireless Wire kit (a pair of pre-configured RBwAPG-60ad running latest long-term, 6.49.18 at the time of writing) deployed on a remote location to extend the service from a 4011 to a CRS326 in the net building. The distance is about 30-35 meters and for the first several months everything was working as a charm. Several weeks ago customers on the CRS and CRS-attached APs started complaining about connectivity losses and the problem always was the wireless link not working.

Digging into the logs there were occasional up-downs of the link:

oct/04 13:04:41 interface,info wlan60-1: link down 
oct/04 13:04:44 interface,info wlan60-1: link up 
oct/04 19:33:49 interface,info wlan60-1: link down 
oct/04 19:33:53 interface,info wlan60-1: link up 
oct/05 12:04:07 interface,info wlan60-1: link down 
oct/05 12:04:07 interface,info wlan60-1: link up 

Apparently only some of the up-downs were causing the link to hang.

Bumping the log level to “interface,debug,!packet” didn’t reveal any extra info on link up-downs, but every link down event was 3-4 seconds maximum and everything seemed to be supposed to work, including the “/interface w60 monitor 0” on the master, reporting:

             connected: yes
             frequency: 58320
        remote-address: F4:xx:ww:qq:yy:zz
                tx-mcs: 8
           tx-phy-rate: 2.3Gbps
                signal: 80
                  rssi: -56
             tx-sector: 26
        tx-sector-info: right 11.4 degrees, down 3.8 degrees
              distance: 33.44m
  tx-packet-error-rate: 0%

So link is up, the master sees the slave’s MAC but traffic is not flowing. Attempted mac-telneting to the slave – no go. A reboot seemed to help and a eventually I found out that a “disable-enable” interface on the master also clears the issue. I’ve scripted a watchdog on the master:

/tool netwatch
add down-script=" /interface w60g disable 0; :delay 1s; /interface w60g enable 0; :log error \"reset W60\"" host=<IP of the CRS>

This seems to work, but in the worst-case the CRS users are sitting without connectivity for 1 minute and that is not good.

Has anyone seen such a behavior and, if so, how did you fix it?
Also, is there any way to get more in-depth logging on the W60 interface ups and downs? I presume the ups and downs are due to frequency change but would like to confirm that