It appears that the core issue in our case is that the AP is not sending ARP packets to the OSX client. We finally discovered this while running a packet capture on two OSX clients connected to the same WLAN on the same AP. One client would see a constant stream of ARP communication while the other client would not see any ARP traffic except its own. The end result is that if the router (Brocade/Ruckus switch in our case) for that WLAN had a default ARP timeout of 10 minutes then every 10 minutes the OSX client would lose network connectivity for about 2 minutes until the client was able to exchange ARP packets with the gateway and then traffic would flow again. Since TAC still hasn't been able to resolve this, I have increased the ARP timeout on all of our routers to 22 minutes (ip arp-age 22) which seems to be the minimum amount it can be set to so that the OSX clients do not drop offline. OSX seems to generate its own ARP traffic at least once every 21 minutes which then avoids the problem. However, the core issue remains that two identical clients running packet captures see very different ARP communication from the APs. Also, if the client roams to another AP and then back to the original AP the client seems to get all the ARP traffic for a while then it goes away at some point.