Keith's auto-response document RFC 3834 actually uses 'MAY':
... a responder MAY ignore any subject message with a
List-* field [I5.RFC2369]. ...
This brings up the point of how closely should the vacation extension
follow the recommendations in 3834? Should we reference 3834 much more
strongly as the basis for how to handle vacation responses, and make
that we use the same MUST/SHOULD/MAY behaviour that 3834 describes?
IMO we would need a very good reason to loosen any requirement 3834
Being even more strict is not a problem compliance-wise unless that
strictness causes technical problems of its own.
In the specific case of MAY vs SHOULD on list- fields, I think a SHOULD
is more appropriate than a MAY.
Personally, I would prefer SHOULD as well.
I did a comparison of the specifications some time back and we were
the time. This was before RFC 3834 was finalized, however, so another
comparison is probably a good idea.