As for the problem of inappropriate bouncing of mail - I do think it
might be useful to write an RFC documenting dubious practices,
such as refusing mail with a null return-path, expecting the mail
to come from a client machine with the same name as the sender's
domain, trusting blacklists, etc.
See RFC 2505 - "Anti-Spam Recommendations for SMTP MTAs".
some of the necessary information is buried in there, but not all of it.
also, times have changed a bit, and I'd make slightly different
recommendations today than are in that document, based on recent experience.