I've been away for an extended amount of time, and finally had a chance to look into this a few days ago. I was able to figure it out with a well rested mind. The problem was two-fold: the Brocade was tagging VLAN1 whereas nothing else on the network was, and there was a port channel configuration issue on the Brocade side. Apparently you can create the port channel from either switch, but it's switchport members need to be configured individually and identically on each switch.
Have you worked out your packet loss issues yet? I was getting a lot of event ID 4230 "TCP/IP has chosen to restrict the congestion window for several connections due to a network condition" errors on my RDS server in VMWare, and thought that it might be the MTU size, but after digging realized it was a network issue to an older server with an old NIC where the user profile disks are stored.
I'm not so sure that your is the mini jumbo frames issue either. What is the MTU size of your virtual NICS? Which virtual NICs are you using? I may be wrong, but if all of you endpoints have the MTU size of 1500 and only the Brocade VDX in the middle is set to 2500, then I can't see why the packets would become fragmented. The RDP packet size is fixed. It wouldn't grow when it leaves the 1500 pipe into the 2500 pipe. Why not change the VDX MTU to 1500, or are you using it for FC and LAN? I'm not clear whether or not the 6720 allows per port MTU settings, or is it global. I know older VDX firmware with layer 2 mandated a global setting only.
What are your physical port stats like with sh int tengig x/x/x or sh int stats x/x/x br? Are there any errors or port discards on rx or tx?
Cheers,
Brad