There is a reason the best practice is not to do huge PRs because it's very difficult to review. Now imagine one person has to review all the code generated by a company such as Twitter with 3k+ Engineers across multiple years. Is that even possible given the cognitive load? If so how long would it take to produce anything meaningful?