Author here:
Spot on, there are so many little things that we as humans use as subtle verification steps to decide how much scrutiny various things require. LLMs are not necessarily the death of that concept but they do make it far far harder.
There's so much more than "works well". There are many cues that exist close to code, but are not code:
I trust more if the contributor explains their change well.
I trust more if the contributor did great things in the past.
I trust more if the contributor manages granularity well (reasonable commits, not huge changes).
I trust more if the contributor picks the right problems to work on (fixing bugs before adding new features, etc).
I trust more if the contributor proves being able to maintain existing code, not just add on top of it.
I trust more if the contributor makes regular contributions.
And so on...