It could be argued (or shot down <g>) that the truly conservative
approach would be then to add a header
X-OriginalSubject: <raworiginalsubject> here
so the user has a HOPE of recovering it if the MUA munged it
accidentally.
surely it's easier to say that MUAs shouldn't mung subjects
accidentally?
(actually I've thought for a long time that it would be useful to have
a Subject-Was: field, that would automatically include the Subject of
the message being replied-to if the person composing the reply changed
the Subject from that of the message being replied to.)