Raised here (I didn't check canny first!) https://github.com/facebook/docusaurus/issues/6998

Yes this is a technical limitation.

The broken link checked actually render the pages to collect the rendered links and compare against the routes.

This is not ideal because it fails too late in the build process and the feedback loop is slow. However it's also quite reliable because it works for any links, including paths that are created using JS (ie not statically analyzable).

We should find a way to provide a simpler broken link detection algo that would be less reliable but that could be faster