Indeed, the world is not IID. As such, test cases should not be a uniformly distributed sample of some mathematical distribution.
Right! And even if it were, in the sense that that's what we should expect as real world input, it wouldn't generally be the best distribution for finding bugs.