I was looking through some of the top test failures in wifi_matfunc, and I noticed that network_WiFi_TDLSPing still has quite a few failures:

https://wmatrix.googleplex.com/wifi_matfunc?hide_missing=True&tests=network_WiFi_TDLSPing&days_back=20&releases=61

---

Notably, many devices (or the APs they are paired with in wificells) don't actually support TDLS, but those will get TestNA (yellow).

Also, there is a long-standing bug ( bug 699735 ) to track some failures distinct from the $subject one.

---

To the $subject failure: this test currently assumes that with a clear cache, it can
(a) perform an ARP lookup request that fails
(b) wait 1 second for a request/reply to complete and fill the ARP cache
(c) perform another ARP lookup successfully

Step (c) fails relatively often (~10% on a particular Link device). Looking into the net.log, it seems that shill is complaining in step (c):

"ARP lookup is still in progress"

So...it seems reasonable to put this in a retry loop, to give it a little better chance at succeeding. Initial testing shows a ~0% failure rate for this test on the same link after retrying another time or two.