Load Test revealed some surprising results

I've got an application that utilizes several components inside of it. At the most basic level, the backend servers turn around and talk to the F5 again. We have started doing some load testing and are seeing odd results. We're running a 6900 platform on 11.2. The first set of tests went fairly well, load was distributed fairly well to the 6 backend servers. On the second run however, the test was ramped up to roughly 5000 concurrent users, and all of a sudden the traffic shifted over to being thrown at one serverm like all load balancing logic went out the window. I have a theory on this, but need someone that can help out with the flow a little better to check my logic. The first URI the client hits has a pool with 6 members. The client gets a 302 returned that redirects them to another VIP for login. This VIP has a pool that has 6 members in it. The client machine then gets another 302 that redirects it to ANOTHER VIP, with a pool containing one member, for a centralized login. After the user logs in, they're sent back to the second VIP where they get a session ticket, and then back to the first VIP to actually login. The backend servers are also doing some trickery where they will reach out and hand session tickets amongst themselves validating sessions and keeping things in check. I feel like the pool with the one server is where the issue is. Cookie persistence is set on all VIPs, so I think the pool with one server is sticking to one backend server, and the F5 is sending it back to the same server over and over again because it sees the cookie. Does that logic seem right? I also am curious if we exhausted the persistence table, or SNAT table, or some other table, and at that point the F5 gave up trying to be smart and shotgunned traffic out. Any thoughts or help is appreciated!!