> The onus should really be on the author to manually verify the mass changes.
I disagree with that, The author often does his best but might overlook the obvious. As mentioned, it is OK if only random samples are reviewed.
I also somehow disagree with that. While the author *must* manually verify the change, it's too easy for the author to miss something after a long work they've done. Being too involved and confident into something make usually people less careful and more likely to make small mistakes, that's somehow in human nature. Another external fresh pair of eyeballs on the change will see things that author can overlook.
In a perfect world, another review would be mandatory for every change. Some reference studies about software quality have measured the code review is almost always profitable in the sense that it detects a lot of bugs for a cheaper price/time than just skipping the review, shipping the bugs, and spending more time troubleshooting them later.
