Thanks Aaron for your response. This requirement comes up in WS-AT (Web Service Atomic Transaction). Here is a more detailed description of the scenario.
A weblogic app starts a transaction on node1. It then invokes a dotNet web service with a transaction registration service url embeded in the soap message. This registration url points to web app running on both nodes of the wls cluster with its host name set to the BIG-IP. Moving on, the dotNet web service extracts the registration url and needs to make an hppt/s request to it to enlist with transaction coordinator on wls.
So here is the two key points. First, the registration request from dotNet is an initial request to BIP-IP. Second, since the transaction initiated on node1 is pinned to node1, the registration request from dotNet has to land on node1. Otherwise, a transaction not found error will occur.
My idea is presented in my initial post. If it works, it will probably be the simplest and most efficient way of implementing this scenario. Now to your other questions.
The session Id will be appended to the end of the registration url using URL rewriting. So a registration service url could look like this:
http://BIG-IP-HOST/WSATRegService;JSESSIONID=NODE1SESSIONID
The session id is a constant for a given node but they are unique amongst cluster instances. It never changes over server reboot.
Thanks again for your time.
Siming