Which reinforces the point I was trying to make - enormous phenotypical diversity can arise from a small and stable genotypic pool.
However, this phenotypical diversity is limited by the genotypic pool and dogs in all their wonderful variety do not give rise to cabbage in one or one million generations.
Again the issue here is the facile nostrum that amoebae can become, given enough time and interesting enough circumstances, elephants.
Dogs in all their variety have an irreducible dogness which does not, by chance, become cabbageness or armadilloness.
You have no idea what dogs could give rise to in a million generations. We've only been playing with them for a few hundred. And in that time, we've produced something that's convergently evolved with a tailless rat (chihuahuas). (You can tell I don't like chichuahuas). A dog won't give a cabbage in a million generations, but it could, IMO, give rsie to something we'd be inclined to class in an entirely different family, if we didn't know its origin.
Also, dog diversity is not limited by the pre-existing genotypic pool. It made good use of mutation.