If you're reasonably comfortable with OO, another way to do it is by separating each test out into its own module, instantiating each module as an object, and then calling $obj->run_test() on each. This is actually very similar to my post on Tasks and TaskManagers which I expanded upon a bit in another post the same day.

I use this paradigm to set up tasks which I can really easily mix and match. I can easily whip up a script that calls different sets of tasks - a common example is where I want to call a subset of tasks from my main script so I can just test those tasks. By making each task as independant from one another as is reasonably possible, this becomes really quite trivial.