I run iOS public betas. The staggering improvement in quality between beta 1 and beta 7 seems to be evidence that they doing largely effective QA of major “happy paths.” What I do find interesting is that in early betas, things are broken that imply there are either constant fundamental rewrites going on, or that architecture is sloppy. (Hope it’s the former.) After the final betas, the remaining bugs seem to be untested edge cases. Of course, with complexity, there’s a combinatorial explosion of edge cases to test, so it’s likely their QA is constantly being outpaced.