Hah. I knew it!
Anyway, there's a bit of discussion at pgsql-general about this, and was talking with Eloy about what to do with these funky indexen. If the pg hackers insist that there isn't a runtime workaround, Eloy proposes a soft "hint" in the XMLDB reports page. I like the idea.
http://archives.postgresql.org/pgsql-general/2007-09/msg00121.php
(And if later turns out that we have lots of special index things for different DBs, this may become a series of extensions to xmldb. For now... keep it simple.)