My take: A scenario to explain the SharpenYourPencil #6 (from page 104) question-
Imagine we have a Stateless session bean called "ClientTotal" bean, which finds out (somehow) the number of clients present in the specified network (the network will be supplied to the bean in the form of an argument to the bean methods, either a class 3 IP or the domain name). All this querying and calculating takes some amount of time. There is a memory shortage on the server (assumption), and the container is able to create and maintain only one object of the ClientTotal bean in the bean pool.
Lets imagine Client 1 calls the ClientTotal bean and asks it to calculate the number of clients in the Java-Developer domain. While the bean is making the queries, another client, Client 2, calls the bean and asks it to calculate the number of clients in the Java-Tester domain. Since, there is only one bean in the pool, and since its in the middle of a method (a business method), Client 2 will have to wait until the bean finishes its task with Client 1. Once the bean finishes with the calculations and services the Client 1, it can go back and service Client 2.
So, there will be a stateless session bean existing per client, for the time period that the business method takes to complete.
Please let me know if I am stating something wrong here.
P.S -> Great work, Marc. Though I am a little behind you guys (the Sark study group), I promise to catch up in the coming days
So, if you find the answers or questions, coming in a little late, please adjust.