Will it be satisfactory if I include weasel words to say that
"the MIME standard [ref] says that gateways may nuke non-Content-*
headers; this should be considered when choosing names for headers
that might be used in a difference= parameter"?
Then the restriction will be gone if MIME changes its mind.
(the rule is silly, IMHO, because any gateway that can preserve
arbitary Content-* headings is probably able to preserve *any*
arbitary heading, while those that can't do an open-ended set of headings
won't be able to handle Content-* either. See the MIME-MHS documents
for one example of trouble)