Milter-greylist

In my typical setup, email from "our" users flows from an internal mail server to internet-facing relays and finally to remote recipients. Opposite is also true, and the incoming relays primarily do 'cheap' antispam filtering, including a milter-greylist (MGL). The two roles are not necessarily held by the same relays, but often the incoming and outgoing ones are the same, and MGL is invoked for outgoing email anyway (even though then it falls through quickly by trusting the internal hosts).

So, this gives us an opportunity to set up an 'inverse' autowhite entry - when "our user" initiates communication with some new remote email user, he or she posts a legit message, and we expect an answer to arrive later, likely from the recipient's domain/subnet or mx/spf named hosts. The guess in general can be wrong, but for many setups and cases it is likely that we can adequately estimate where a response would be coming from. 

Thus we might add an auto-white-list entry in advance, so that the initial communications with a new counter-agent are not lagging for the first message exchange - that can cause grief and annoyance. 

The same technique could be (ab)used by "our users" who expect a message from somebody which is in fact delayed by our greylist (i.e. "our user" learns by phone that it was posted an hour ago, but has not fully left the remote sender's relay due to our MGL and so hasn't arrived to our mailbox yet). In this case users often begin testing what's wrong - by posting messages to each other... and thanks to the proposed inverse autowhitelisting, this could just unlock the greylist for the expected legit mail.

I hope my use-cases serve well enough to explain what I want ;)

The question now is whether this functionality exists already, or how it might be best be implemented (and in a perfect world, including also a way to participate for outgoing-mail hosts that do not have a milter-greylist running)?

Thanks in advance,
Jim Klimov

--
Typos courtesy of K-9 Mail on my Samsung Android