...in the last few days: at first most of my rigs suddenly started to error out almost all WUs (code 1 (0x1, -255)). I wouldn't have been surprised about this if I had changed anything on them...but I didn't. They were running without any modifications (e.g. oc'ing) for some days/WUs. What I did was setting network activity from 'suspended' to 'always available'...but I can't believe that this could cause such errors. Now I switched the video cards between two of them and yet it seems that they are back to normal again (but I suspended network activity)...
Then I had a look at my 'oldest' rig lately: the last WU was successfully submitted at 02:03...and 6 hours later, when I wanted to send another two finished WUs I got the message: 'Client detached'...
What ? I haven't done anything to that rig in the meantime... ???

Yep, for sure... ;)
But I changed cards after the rigs started acting stupid...
And my 'old' rig kept its cards...

Oh, I see...
Had a reboot before changing cards...and the rigs still acted crazy...
Anyway, I hope that's history now...

I got those type of error on earlier versions of the BOINC client, that when I set it to "always allow network access" but the network failed, it chrashed almost all units. Some projects were more resistant to this than others, so they kept running while the others errored out.

Now it is better somehow, even when my wireless fails, everything is running fine, no errors.