Two nodes W-123-1620  La R and 0-63-Ridgemount lost there ability to link. Further in my investigation to determine why these to boxes stop meshing I use the command sigspy.  The  result of using this command showed increase in noise between nodes.  Then the noise stopped and the connection came back.  The setup of both these nodes are a clean line of sight.  The area showed over 30 different wireless channels being used.  

The area is densely populated which makes me wonder if all these 802.11 devices can really intermix?  I know some pundits had once said that even leaky microwave ovens can make noise.  I wonder if it is something more like white noise.  A machine like a generator ?

Blue Arrow shows the missing link.

You may wonder why I don't think it is a hardware or software problem.  Because in all cases if I take the equipment back to lab environment everything will work.

The link between these two points has been poor. I changed out equipment and antennae many times.  To solve the problem I changed the antennae to be directional; although, this improve the quality of link both units; this restricted other client from getting on to the network.  The omni direction not only provides a link to other nodes and to the gateway but they also provide connection to the client site.  Changing the radio pattern then affects the ability for the client to communicate with these nodes!

I will look into getting a splitter in order to put in directional antenna and omni antenna on the same node.  It may help.