How do mail servers decide when and where to send "out of office"-type autoreplies?
It seems this is mail server dependent, so some will use a reply-to: address, others the From: and perhaps the Sender: address. Some servers (or all?) seem to care about the Precedence: field and will not send out of office replies at all on Precedence:bulk (or maybe precedence:list?). The return-path seems to be used for more severe types of bounces, but maybe it's used as well?
What I want is to achieve (this time) is to minimize the number of response e-mails that go to the from: address. I don't mind getting them at the reply-to address.
I've been looking for some sort of reference or best practices - any references or suggestions? I 've looked at some relevant RFCs (RFC2822, RFC3834) but I'm looking for how things are actually handled.