Also, instead of commenting on specific files you can comment on whole pull request. That way it wont get marked as outdated. Using pull requests for discussions is a good practice.

Speaking to empirical observations (not to how it works), this isn't entirely true. Comments on specific lines are not outdated when those lines aren't changed. I literally just went through all the comments in a PR after a push and all the comments on specific lines were retained (not outdated), but every file comment was marked as outdated.

In our code reviews, we create issue comments on lines. Those issues are addressed and changes pushed. Comments attached to changed lines get outdated, but those not on changed lines remain.

I don't claim to know the inner workings. I'm just speaking from lots of experience using PRs for code reviews.

And when comments pertain to a file but do not have relevance to a specific line, like a discussion about why the file was added or removed, then attaching that comment to the file and not a particular line makes a lot of sense and I would argue is good practice - better than what we used to do before the ability to comment on files was introduced.

Lastly, commenting on the PR is inadequate because issues brought up there do not get a "resolve" button.

It's inane. I've gotten into the habit of commenting on the top line instead of using the file comment feature, since doing so almost always creates a lot of repeated work transferring comments to retain them. It's less than useless the way it is now. It actually creates braindead work.

That's the exact problem I'm experiencing; having to recreate comments on files after a push! It incentivises just using top-line comments (which is exactly what we were doing before this feature was added).

I appreciate why it works the way it does, based on @nickkipshidze's comment, however there must be a workaround, in its current implementation the feature is useless-adjacent.

But if that's how it works, then why are comments on lines of a file retained when that file is changed in a push to the PR (unless the comment is on changed lines)? Based on @nickkipshidze's logic, it seems like all comments on a file changed in a push should be outdated, but that's clearly not the case. Why don't the file comments work the same way we observe with comments inside the file?: retained unless the name of the file or its location changes?

I mean, why would commenting on the top line of a file produce a durable comment that persists between pushes if comments are attached to commits and I pushed a new commit?

I'm sure there's probably a technical explanation about the difference between a comment on a "line of an unchanged file" and a comment on an "unchanged file" that explains the behavior difference, but if there is, it doesn't change the fact that it makes comments "on files" more difficult to follow up on compared to comments on the top line of a file.

Top line commenting is not possible for submodule changes & diffs large enough that github refuses to render them. To comment on such changes in a PR you need to use the "file" comment, and it is so very incredibly annoying.