When you say "RMI protocol feature", I'm assuming you're talking about the tool, rather than Remoting. Remoting has an RMI transport, but, by default, EJB2 in JBoss AS uses the Remoting socket transport, so that's what I'll address.
A Remoting server using the socket transport doesn't see clients, per se. It has a pool of worker threads, each of which manages a connection to a client. Now, on the client side, multiple clients running in the same JVM and connected to the same server will share a connection pool, where each connection in the pool is affiliated with a distinct worker thread. If you have 50 clients running in a single JVM and they never do more than one simultaneous invocation, there will be only one connection in use and only one worker thread for all of these clients. But if a client makes an invocation and the single connection is in use, a new connection and a new worker thread will be created. If the clients are very busy and make 50 simultaneous invocations, you will have 50 connections and 50 worker threads.
Now, if your 50 clients are all running in different JVMs, there won't be any sharing of connections, so they will each have their own connection and worker thread.
Hope that helps.
Ron, thanks for the quick response. At least now I have a better idea of what will happen on the server and should be able to talk a little bit more intelligently with our Performance Engineering group.