The problem with this is how to decide which students fail and which pass a course. If you just use a cutoff on the z-score you will be consistently failing a certain proportion of the students, which might sound reasonable but completely ignores the fact that one is supposed to pass only the students that display knowledge on a minimal amount of the material. Theoretically, and A student completely knows the material, a B student mostly knows, with some less important parts missing, etc, and completely relativizing grades will lose this validation.

Another problem is that being a slightly-less-crappy student in a crappy class will give you a better z-score than being a great student in an excellent class.