I have no new suggestions on how to handle "old" FITS files
with multiple keyword occurances, but we might consider
a convention to reduce the problem in the future.
Such occurances probably occur mostly when histories of
previous reduction steps are copied into the headers
of downstream products. Originally this kind of
information was supposed to be recorded with
HISTORY keywords, but I can imagine that many
programmers ignore this option because it requires
reformatting each card image. Maybe something like
BLOCK HISTORY BEGIN
.... literal copy of keywords from previous steps ...
BLOCK HISTORY END

would be useful.

Walter