This seems like a particularly limited view of what a machine is. Specifically expecting it to behave deterministically.

Then you are going to hate the future.

Exactly this.

There's a lot of shit that has become "best practice" over the last 15 years, and a lot more that was "best practice" but fell out of favor because reasons. All of it exists on a continuum of what is actually reasonable given the circumstances. Reviewing pull requests is one of those things that is reasonable af in theory, produces mediocre results in practice, and is frequently nothing more than bureaucratic overhead. Consider a case where an individual adds a new feature to an existing codebase. Given they are almost certainly the only one who has spent significant time researching the particulars of the feature set in question, and are the only individual with any experience at all with the new code, having another developer review it means you've got inexperienced, low-info eyes examining something they do not fully understand, and will have to take some amount of time to come up to speed on. Sure they'll catch obvious errors, but so would a decent test suite.

Am I arguing in favor of egalitarian commit food fights with no adults in the room? Absolutely not. But demanding literally every change go through a formal review process before getting committed, like any other coding dogma, has a tendency to generate at least as much bullshit as it catches, just a different flavor.