Re: second attempt

You could use the various hacks
you've discussed, with modifier letters; but that is not "encoding",
that is "abusing Unicode to do markup". At least, that's the view I
take!

+1

In general, I have a certain sympathy for the position that there is no universal
answer for the dividing line between plain and styled text; there are some texts
where the conventional division of plain test and styling means that the plain
text alone will become somewhat ambiguous.

We know that for mathematics, a different dividing line meant that it is possible
to create an (almost) plain text version of many (if not most) mathematical
texts; the conventions of that field are widely shared -- supporting a case for
allowing a standard encoding to support it.

However, it stops short of 100% support for edge cases, as does the ordinary
plain text when used for "normal" texts. I think, on balance, that is OK.

If there were another important notational convention, widely shared,
reasonably consistent and so on, then I see no principled objection to considering
whether it should be supported (minus some edge cases) in its own form of
plain text (with appropriate additional elements encoded).

The current case, transcribing a post-card to make the text searchable, for
example, would fit the use case for ordinary plain text, with the warning against
simulated effects of markup. All other uses are better served by markup, whether
SGML / XML style to capture identified features, or final-form rich text like PDF
just preserving the appearance.

A./