This is misapplying the unix philosophy. There would be too much duplicated functionality between different tools for different sites. It would be like having different versions of /bin/ls for different file systems.
Youtube-dl should be architected to allow many other people to own the site-specific support (i.e. via a plugin system, with the main "distro" of yt-dl including some popular plugins, which other people or themselves are responsible for), if it was like that then probably this fork wouldn't be necessary.
Another advantage would be to easier maintain those extractors as well as all issues that arise around specific ones. So no more thousands of issues addressing different extractors in one project.