Literally none of those is a situation I've ever encountered. I don't think it's as hard to avoid as you think.

All depends on what you're doing. When I worked in computational linguistics of languages that don't use ASCII (or European "accented" characters), I was very concerned with the display of non-ASCII Unicode scripts: Arabic, Devanagari, Thaana, Tamil, Chinese, even at one point Cambodian (which was extremely hard to render, even in XeLaTeX--it's improved since then). But at the moment, accented European characters are all I need.