> I think the sleazes I specify for patching CHAR-{UP,DOWN}CASE are
acceptable,

I agree.  I meant, after subsuming all of your sleazen.

> Heh? What about Latin-1's es-szet? Decomposing turns it into "ss" for
sure?
> Or leaves it as es-szet?

As I read the specification, sharp-s does not mechanically decompose
into ss.  I was conflating the recent German spelling reform with
decomposition.  My mistake.

> What is done with downcasing Unicode's position-sensitive Greek
letters?

Lowercase sigma is the only such Greek letter.  Follow the official
Unicode data-table, which dictates that U+3A3 maps to U+3C3, i.e.,
non-final sigma.  That is correct at the character level, though not at
the string or text-block level.  Initial/medial-sigma and final-sigma
are the same letter, just written differently.  (This should be a
shaping issue, like writing Arabic, not a discrete character value.  But
it isn't.)