Table timeout/lifetime Not Reflected in -count

So, there are several things that could be going on here. Without typing up a whole diagnostic tree to go down, let's see if we can eliminate one of the most common problems we've run into which could be causing this behavior. Specifically, this behavior could be evidence that you're hitting ID 247874, which happens because there is some problem with the HA channel between boxes. What version are you on? By any chance is your "controlled environment" BIG-IP configured as a standalone box, but your "production environment" BIG-IP configured as HA/failover? If your production box is set up for HA, then please double-check all of the HA configuration, especially peer IP addresses. If the problem remains, then the most definitive way to test for that ID is to set the "StateMirror.MirrorSessions" DB variable to "disable" on the boxes demonstrating the problem and then rebooting them. Perhaps you could try that during a maintenance window. If the problem goes away, you are definitely hitting ID 247874.