Extending protein alphabet with non-standard amino acids and DNA/RNA alphabet with additional "letters"[1] could drive an explosive growth in the amount of possibilities.
As an analogy, it seem a little like a programming language - up to a point, adding more keywords and operators makes it more 'powerful' (all Turing-complete languages are equally powerful, yes). However beyond that you are just adding more specialised parts that could achieve the same thing with more, simpler parts.
I guess I am saying that the leap from, say a 5 amino acid alphabet to a 20 letter one might result in an explosive growth, but it is diminishing returns after that.