- Does it fix the problem without introducing regressions?
- Does it have tests?
- Are the docs and code style OK?
- Does it remain within scope?
- Is it the best fix we can come up with?
And then she goes into loads of awesome detail, such as that interface changes need browser testing and screenshots, reading backwards first to get the code style and comment issues out of the way before thinking about the quality of the code, and using the core gates mentally.