Well, you're actually thinking in the right direction. Designs that efficiently incorporate many categories in game title-based clusters have been proposed
as early as 2006. In fact, even at that point most people who had discussed that agreed it was an interesting idea worth implementing, but unfortunately it required reworking a significant amount of ways the site did things, and the snag that prevented its realization back then is the same these days: lack of time (and probably motivation) to do a proper job. Now that the site has new administration, it's up for them to allocate site coding resources and negotiate with resident coders with regards to access to the site and adding new features, and this doesn't seem high on the priority list for either.
The idea never lost its relevance, though, because severely limiting the amount of categories is a persistent problem that hurts the site in many ways. The "we don't need many categories" mentality lost its charm back when standards of TASing rose to a point that makes subsequent improvement tiring and completely unwelcome for beginners. I.e. back in 2007.