A number of different "standard tools" on the shared Linux machines (e.g., Netscape, tcsh, java) routinely go into infinite loops, sometimes with no obvious sign that this has happened (e.g., they loop while attempting to terminate, but after the window has gone away). This means that without some kind of intervention, the shared machines end up with a large number of (furiously) running processes, slowing response for everyone trying to use them.
We've handled this manually in the past. The intent of the this policy is to handle it automatically. The CPU consumption limits are set much higher than experience shows processes in active use consume on those machines, the way we use them. So, overwhelmingly, exceeding even the 1 CPU hour limit is a sign that something has gone wrong.
It is entirely routine for users on these systems not to know that they have such orphaned processes running, and not to know how to terminate them. The mail sent at 1 CPU hour simply lets them the know the process is running, tells them how to kill it if they weren't intending for it to continue running, and tells them what to do if this was intentional and they think the process might go over the 10 CPU hour limit.
The mail at 10 CPU hours is sent after the process is killed. 10 CPU hours is intended to be long enough for "overnight runs" by interested students doing things beyond coursework.