I took the request for documentation to be rhetorical. If there had been serious accidents using AP2 on local streets we would have all heard about them. The happy fact that there probably haven't been doesn't mean the system isn't dangerous. The reason, in my opinion, is that those of us who try it are hyper-vigilant, actually working harder than we would be if we just did the driving ourselves, and correcting the errors before they can cause crashes.
As a member - and presumably Tesla owner - since 2014 your perspective is different from mine. You early adopters were pioneers, and I applaud your spirit of adventure. Buying in the last quarter of last year I didn't sign up to do stress testing. I bought a highly rated production car with a set of features that were claimed to be in "final validation", with expected delivery by year end. That claim was false, as we now know that a required long software development process was just beginning when the claim was made.
So, having been victimized by a cynical marketing decision, some of us are somewhat cynical ourselves. Tesla and others have twisted the meaning of "beta testing" away from its original concept. Beta testers are volunteers who elect to try systems that aren't fully functional, either for the fun of it or out of altruism. When Tesla says a feature is in Beta, they simply mean it doesn't work well but they decided to put it into production anyway. Local street AP is an example, and I think it's irresponsible for them to assume that every driver will understand the limitations enough to keep himself, pedestrians, and occupants of other vehicles from grievous bodily harm.
