So there seems to be some confusion around how the RPC Proxy component on an Exchange 2010 CAS server in a CAS array behaves when it comes to incoming Outlook Anywhere connections.
When using a CAS array in combination with a hardware based load balancer (HLB) solution, will the CAS server forward the RPC traffic back to the VIP used associated with the RPC virtual service on the HLB solution?
Something like this:
OA > HLB > CAS array > HLB > CAS array > MBX
Fortunately the CAS server doesn't forward the RPC traffic back to the HLB solution.
Windows Server 2008 and 2008 R2 includes a component called LBS (Load Balancing Service) leveraged by Exchange 2010. In regards to incoming Outlook Anywhere connections, the RPC Proxy is designed in such a way that it initiates a callback to LBS which then ensures that the CAS server handles the request instead of forwarding it to the HLB (to which the CAS array FQDN resolves).
MCM: Exchange 2007 | MVP: Exchange Architecture
MCITP: EMA + EA | MCSE: M + S | TechNet Influent