>However, given the existence of diverse populations assuming the Model A, we can conclude that these populations have found a local minimum of fitness in which diversity is advantageous.

But any mechanism in which "diversity" is advantageous will be advantageous for more specific reasons. Diversity is just too unspecific. Perhaps you're speaking in terms of group selection, so in a sense evolution evolves diversity in populations. But I don't think this works either. Aside from group selection itself being highly questionable, every environmental event in your Model A would reduce the diversity of the remaining populations. Remaining populations were more diverse in the past (owing to their ability to survive abrupt environmental shifts), so any trend in diversity is in fact downward on large timescales. On short timescales, diversity only grows when there are no selective pressures at work on a given gene. There just is no pro-diversity selection going on.