Well, in priniciple variant specific application has nothing to do with CRISPR, any genome alternation could be coupled to certain mechanisms that rely on reading out the specific traits of the targeted humans. Of course, the current developments enhance these possebilities with gene drives etc. But on the other hand it also allows the development of defense mechanisms, no technology is the ultimate game changer (though CRISPR is thought to get pretty close).

Spread through food or water supplies is probably possible, given that CRISPR is currently mostly administered via virusses like AAV or lentivirus. But that nothing new, those virus cassets existed before, and can also be used with other methods like zink-finger nucleases or TALENs.

Delayed? I don't know, probably. It really depends on what you want to do. Keep in mind that most current applications affect around 0.5 percent of all hit cells. The effectivness of the process heavily depends on the construct and probably decreases with complexity. But sure, delayed promotors and other regulatory circuits should be possible, though unlikely for now.

The whole technology is dangerous, so dangerous that we probably cannot stop it now and have to form and regulate it. But it also has enourmous benefits associated, and again, can probably defend against itself. You cannot win evolution, only participate.