I may be experiencing this, not 100% sure though. We're doing a phased upgrade for iOS and all the Apple apps by updating one at a time over the course of several days. I kick the process off at the end of the day to avoid bogging down our public connection. I'm seeing maybe 40% of my update traffic hit the caching servers and the rest is going out the Internet. I'm not sure on the exact percentage since all 1.2 Gbps of my Internet pipe gets chewed up in the process.
We use 4 public IP addresses for our network and all 8 caching servers are configured to server all internal clients. They see each other and all of the settings look correct. Two of the public IPs are the starting and ending addresses of the range, so I'm kind of wondering if the setting isn't inclusive? I'm waiting to hear back from the Apple engineer we have assigned to our account.