So, I was reading the related question here, when I went off on a tangent.
The point of the given post, as I read it, is that certain languages like GolfScript, K, and some other extremely-high level languages can trivialize Code Golf by taking problems (for the sake of discussion, determining whether some string is a palindrome), and creating a solution in likely less than 15-20 characters, which totally blows other more verbose languages out of the water; they simply can't compete!
A potential solution I want to discuss is probably not a real shocker, and might seem a tad obvious: winners can be tallied by language. To return to the palindrome example, say User A creates a 65-character Python solution and a 15 character GolfScript solution, while User B creates a 45-character K solution and a 25 character GolfScript solution. In categories, User A would be the winner for GolfScript, while User B would be the winner for K (at least, until some challenger comes to take their title away.)
The major point I can see against this idea is that there's only one Green Checkmark of Victory. This idea is imperfect because splitting things up by category raises the question of what it really means to win at Code Golf. Dividing by language means the possibility of multiple winners, which is thoroughly incompatible with the format of a Q/A site.
So, the Meta-Question: A) How can we compartmentalize these competitions such that users of arcane, high-level languages can recieve recognition alongside users of more verbose languages, and B) If we accept a multi-winner paradigm, how can we crown the One True Winner?