You can deal with the confusion of a half-indexed corpus of messages,
because you have your hands on the constructruction of the site and know
its structure inside out. Is there anybody else who needs to access the
site? Having indices that work for some messages and not for others sounds
like a gold plated way to convince would-be users that the resource is
broken and unusable. Just a thought.
The site is http://www.mail-archive.com and has quite a few users. I
suspect nobody would notice if I pruned the MHonArc indexes to the
5000 most recent messages. Given the current user interface, do you
think I will alienate users?