You can operate anyway you want if you have a printing press for money in the back. It's not a model to follow when you've got constraints.

Facebook has a very thorough code review process, and their product is heavily data driven. They've even productized their code review process internally, and that product is also data driven.

The code review process is quite literally “you need a stamp from anyone because of some regulations we follow”. Definitely nothing thorough.

The seriousness of any code review process can be determined by whether it diffuses liability from the original code author for any bugs.

So far in my career this has never been the case.

I think we do, to some extend. As the TL, when I’ve reviewed their PR I’m definitely not going to blame them for the bugs I missed during review. They’ll still have to fix them, because that’s just most efficient.