Created attachment 1272 [details]
trace

Can you take a look at this trace, richard?

Richard...some observations from the trace , rgds Dave Williams

it looks like the beaconing from the AP is either not correct when the DSL is connected , or the squeezebox tries to make probe requests when the beaconing goes out of shape. ooks like the beaconing should occur every 0.1024 seconds...keen to hear how sensitive your squeezebox wireless implementation is to delay,jitter or loss in the beaconing.

I think the root cause of the symptoms observed falls between two camps. Draytek router wlan mgmt traffic is impacted by the dsl connection..and the squeezebox seems to be sensitive to the impacted wlan mgmt traffic. 
that's my theory (for now at least)..but grateful for an expert view (which i am not) to help fix.

look at frame 29298 to about 32000 - this is with the DSL connected
from about 32000 onwards the connection is solid with the DSL disconnected

Looking at the trace in ethereal it looks like the wireless card in the Draytek router keeps rebooting. To see this clearly use this filter 'wlan.sa == 00:14:85:bc:3d:a9'. You need to look at the timestamp in the beacon frames.

Notice how this is lower between frames 29247 and 29280. Again between 29436 and 29437, etc. At the start of the trace, and later in the trace the timestamp increments as expected. It is easy to see where in the trace the timestamp is reset, as the sequence number is also reset at the same points.

I have check the squeezebox wireless implementation and it checks if the beacon timestamp is lower than the previous beacon, in this case it assumes that the access point has rebooted and authentication/association is therefore needed again. This triggers the squeezebox to start sending probe requests again.

During the authentication/associtation the Draytek must send the squeezebox a data frame (not captured on the trace), and this results in the dissassociate management frame (29298) with reason code "Class 3 frame received from nonassociated station".

At this stage I am reluctant to modify the way the squeezebox uses the timestamp to detect when an access point as been rebooted as this risks regression with other access points.

Dave I suggest you run this past Draytek support and see if they can fix the router firmware.

Richard

According to the recent posts in this thread http://forums.slimdevices.com/showthread.php?t=10458&page=3 Draytek have released a new router firmware that resolves this problem. If you need any further assistance please re-open this bug.