"does the code work functionally" is a fraction of the overall purpose of a code review. If you rely solely on "does the code work", you end up with the average Rails app after 1 year: 10k line controllers, deeply entangled code, half uncommented code and half deeply over-commented code. Hundreds of comments inevitably end up out of date. Whole blocks of code might be commented out. jQuery is instantiated alongside React and 2 different versions of Angular.

I'll repeat that there's not a single code change I can think of, including comments and documentation, that can safely be merged without a review from a second set of eyes. This is exponentially true the more engineers work on the codebase. I can maybe understand why someone on a team of 3 or less doesn't see the benefit in having the overhead of mandatory code reviews.

If you push into master and forget to run tests (which happens, people are human and make mistakes, especially with small changes), then your broken tests are now breaking everyone's attempts to run the build though. You don't get the feedback that your tests are broken until after it's pushed into master.