I am assuming that when authors choose to use ch that they are aware
of the perils involved ...
I wouldn't assume this. If you want to leave "ch" as a width unit,
then you should point out the perils.
The use of ems is problematic in this case as an em isn't the same
as the width of a character for a fixed pitch font. My understanding
is that the em is equal to the point size which is related to the
height of the font - not its width.
OK, but the problem remains of how to specify what "ch" means in
the context of a proportional font or a mixture of fonts. You could say
something like:
In a proportional or mixed font environment, the meaning of "ch" is
the nominal width of 'M' (or pick your favorite glyph) in the case of
a single font or the average of multiple 'M's in the case of multiple
fonts.
Alternatively, you could introduce the language used to describe the
traditional process of copy fitting in which an average width is computed
by taking a specific sample text in the given font. [This strikes me
as somewhat overly specific in the context of HTML.]
Glenn Adams