I'd prefer to shoot for the curated approach, at least to start. I don't think opening a PR is such a burden.
My long-term goal is that this doc has checked links and so forth, and moves into the repo, such that when people make edits to the compiler that would invalidate it, it is expected that part of the PR is to update the guide.
This may be unrealistic, but it's worth a shot.
As a side benefit, it's much easier to read a "big refactoring" PR when you have an overview of what the author is trying to accomplish.