Haven't read the book yet, but isn't that more like a matter of the font/rendering engine? I have a murky notion that for Cyrillic, for example, there are some nuances in rendering certain glyphs in cursive between languages [1], but these nuances are usually resolved by cooperation of the font and client interpreting the language hints, so not in the "physical" text.

(Not telling I see this as a good thing or anything: it is way beyond my expertise; I definitely can see the motivation for introducing as many variants in the Unicode register as there are in the real world)

Isn't the umlaut vs trema/diaeresis in a similar situation?

[1] made me test it and cobble a demo. (Sadly, not speaking any of these languages, so cannot verify it is correct; just wanted to see the difference in practice):

    data:text/html;charset=utf-8;verbatim,<style>
    @import url("https://fonts.googleapis.com/css2?family=Noto+Sans:ital@0;1");
    body { font-family: 'Noto Sans'; }
    dl:hover i { font-style: normal; }
    </style>
    <dl>
    <dt>lang="ru"
    <dd lang="ru"><i>грипп, практика, график, типа</i>
    <dt>lang="sr"
    <dd lang="sr"><i>грипп, практика, график, типа</i>
    </dl>

Arguably, depending on wide (physical text ↔ specific font ↔ rendering agent) ecosystem feels quite fragile, but cannot tell if there is any better alternative for this particular case.

https://myfonj.github.io/sandbox.html#%3C!doctype%20html%3E%...