Kind of busy right now (grape harvest has started) so I can’t respond to everything yet, but I just wanted to address this point.
The initial total genetic variation within a hundred unrelated individuals is the same as exists in an interbreeding population generated from those 100 individuals. It doesn’t matter how big that population is, there will never be more genetic variation than originally existed in the parents, outside of rare chance mutations. Let’s say you let that population interbreed for 100 years, selecting for drought tolerance. After 100 years of selection, unless you maintain the original parents somewhere (which, given you can clone trees virtually forever isn’t hard to do), you end up losing genetic diversity, since invariably, some combinations of alleles will prove to be less beneficial under selection conditions and some will be lost by chance.
Adapting a population by definition means you are selecting against traits that aren’t useful or desirable for local conditions; i.e. you lose genetic diversity. I’m all for creating landraces, but I’m arguing for preserving whatever cultivars we can. Many of them are the result of thousands of years of selection followed by decades or centuries of vegetative propagation. Cultivars, especially heirloom ones, represent a part of our cultural heritage that we won’t ever get back if lost, even if some of their genes live on in a landrace.