The questions these comments brings to mind are the effects on a three tier mail design a pair of us were playing with once. The idea was the outermost layer (MX points to these servers) just does nothing but ID whether email is talking to a legitimate address. In this case the address book would be some common source, such as a LDAP server. Second layer would accept messages and spam check them, and then the third layer would be the mailer customers interact with.
In an entirely paranoid implementation the outermost layer accepts everything and drops anything illegitimate silently. Whether that violates RFCs aplenty I don't know. It's been a while since I've done email admin. But it would prevent harvesting of addresses via directory attacks.