Dozy Wannabe,
You make some good points about software testing. Having worked in the field of research (Best part of 30 years ago) that was looking at how to prove programs correct for life critical systems (A320 included), one of the facinating topics was how to add the extra dimension of time to unit and regression testing.
If all expected (unexpected and down right surprising) inputs to a function (should say object but let's keep this in layman's terms) produce expected output, then you have a function that works in isolation. The problem with any "real-time" system is that you do not know when such inputs will occur and what else will be happening at the time. I no longer work in this field but the probelm of how to test for an acceptably comprehensive set of time related inputs is something that was not solved at the time.
Keep testing until you are satisfied that nothing new will occur was the method. This can take a while in a system as complicated as that flying the A330. There is no doubt that the "two teams" approach that you describe for the A320 helped bring a lot of confidence to the testing teams.