A bit OT: Cut off address probes?

[Furnish, Trever G]

This is probably a bit off-topic, and I hope it's not a faq somewhere
already - feel free to yell at me constructively if so. :-)

Is there a way to configure sendmail (or whatever) such that "address
probes" are less effective and intrusive?  I could imagine how the process
might work, but I've never coded a milter and am hoping someone else has
done this or will tell me why it would be a bad idea.

By address probe, I mean connections that either:
        1. Ask the receiving mta to accept a message for one invalid address
after another despite repeated negative responses from the receiving mta.
Something that amounts to "Is bob valid?" ... "no"  ... "Well, what about
tom?"  ... "no"  ... "Frank?" ... etc.

        2. Send a message with many recipients at the same server, learning
those that don't bounce.

Completely blocking "no such user" responses seems like a bad idea, but
ignoring someone who attempted delivery to X number of invalid addresses
within Y seconds seems like a good idea.  But how can the first-line MTA
know whether or not an address is invalid?

Has anyone set up or read of such a system?  Perhaps a sendmail milter that
looks up recipient addresses in a flat file or via ldap before accepting the
message?

--
Trever