Werner will review anything in App, Gui, Part and PartDesign, and many in other areas too. He's probably the most skilled of us all, and by far the one with most FreeCAD experience as he's there since the very beginnings. So yes, his authority is unquestionable. But it's not true that he only trust himself at all. All the way FreeCAD grew up to know is based on trust between people.
But you're right we should write better docs to explain what is needed.
The main problem we have really is with large pull requests that touch many files. It's difficult or impossible to measure the impact of such changes only by looking at the code. So what's mostly needed there are people wanting to compile the branch from the pull request, check the functionality that the PR proposes (check that it correclty does what it is intended to do), and, that's the most important and the hardest part and that requires a fairly good knowledge of FreeCAD, give some stress-test in all areas possibly affected by the proposed changes.
For example, a PR that would propose changes to the tree view: What would need to be tested is that the changes are working correctly, but that other operations that use the tree view or redraw it are still working ok: adding objects, removing objects, right-clicking everywhere, selecting, embedding in parts and groups, switching workbenches, try to make possible problems appear.
This helps enormously to make the merging faster, because it takes some time (build,run, test...) and it's hard to automate and humans are far better at causing crashes than automated tests
and if you see that someone else did all this and reports no error you can have much, much more trust in the PR already