I'm not sure I explained it very well, or I'm not understanding your goals correctly. I'm not sure where the description of it as a directory-based approach comes from, but it's not necessarily directory based. You can attach rules to a directory, but you can also put them at the top level and just execute the rule based on a regular expression match. It's really just an engine wired into UpdateMedia that executes custom logic against media that it matches via regular expression. Isn't that what you need?
In your case I was thinking a rule for each data source (e.g., "IMDBRule", "TVDB"), which you could apply at the top level (or lower if folks did happen to have a tv directory they wanted to limit it to) for an array of regular expressions to match the various naming schemes you wanted to support. Any media matching would cause your rule to execute.
Am I way off base here? It still seems like a good fit to me.