link-checker incorrectly says a fragment does not exist

[mcking65]

In #2734, the link-checker is failing on the link:

https://github.com/w3c/aria-practices#code-conformance

It says the fragment #code-conformance was not found. I wonder if it is added dynamically and the link-checker is getting incomplete content?

this test failure is blocking progress on #2734.

[howard-e]

Something interesting I noticed after looking through the page ids exposed by this repo's link-checker solution, is that for a GitHub page, almost every element on the page has an id so it can be navigated to and focused. So https://github.com/w3c/aria-practices#issues-tab or https://github.com/w3c/aria-practices#fork-button focuses onto the Issues Tab and the Fork Button respectively.

That would mean the chance of id name clashes increases with those reserved names, so the ids of the heading elements in the README are actually prepended with user-content-. So for example, user-content-code-conformance. This makes it possible to navigate to https://github.com/w3c/aria-practices#user-content-code-conformance as well as https://github.com/w3c/aria-practices#code-conformance. I suppose GH does some intelligent matching in JS when navigating with the stripped version (as long as there's no clashes with their reserved ids?). But the link checker is explicitly checking for code-conformance.

So I think this means that the check for the "hashes" could be relaxed to something like the following or a bit more safe:

ids.some(id => id.includes(hash));

Would we want to explicitly check for this with GitHub links?