Sergey Organov [off-list ref] writes:
2. We have descriptive long name for every other option, and it'd be an
exception if we'd have none for --diff-merges=m. In fact, it's
--diff-merges=m that could have been removed, but it'd break resemblance
with --cc and -c that both do have their --diff-merges=cc and
--diff-merges=c counterparts.
Hmph, a devil's advocate in me suspects that it may just be arguing
why user-configurable 'default' is a bad idea, though.