I was surprised to hear that the use of one TeX hyphenation trie
for French (say) is incompatible with the use of several font encodings,
say T1 and the Y&Y encoding.
To my mind, the problem had in principle been solved long ago.
Approximately as follows:- Use the hyphenation tables for T1 encoding.
Where the Y&Y encoding is in force, use a hyphenation reincoding vector
(\lccode in TeX, \hyphencode for eTeX) to map from Y&Y encoding points
to "equivalant". Well, we all know that TeX blew it since the linking
between \lccode and hyphenation is assayed only when a paragraph
closes. Strangely, eTeX blew it too (so far) -- by allowing only one
\hyphencode vector *per* trie, whereas in in the example in hand two
are wanted, one for T1 and one for the Y&Y encoding.
Why did the eTeX folks tolerate such a silly restriction? Clearly to
discourage immoral deviations from T1 encoding! Unfortunately morals
like fashions do change from year to year, and then TeXies find
themselves wearing shoes that do not fit.