It appears that surefire and failsafe plugins execute test classes in order while tests defined within a class execute in undetermined order.
To discover tests which rely on order (what we consider bad tests) we want to force the order to be different for each run. Ideally, we'd have a mechanism to disable randomization or a seed number that would repeat the order (must like the old palm OS emulator had a seed number that drove a sequence of random tests).
Let me know if you know a way to do this? If not, I guess I can work one into a local fork and then submit it.
Thanks
Peter