Small follow up below:
Tony Finch wrote:
Exim has a database of which messages are waiting for which hosts, so
when a host comes back to life it can flush the messages down the same
connection without having to wait for the next queue run.
So I suggest that the software is collecting, sorting and probably maybe
related to total counts, like gmail.com addresses and others common
subscription addresses, that it spawn delivery job based on the highest
count first and the single host mail as done last and maybe
alphabetically. I suppose total failure history to connect is a factor,
but our MX servers are rarely ever down.
The only point here is to express another reason why we choose to send
mail out autonomously, equally as they come for the highest efficient
and throughput. When you priority based on counts, etc, the majority
which are count=1 host mail will be delayed far longer than it has to been.
The only time wcSMTP stops and blocks a host domain is with NXDOMAIN or
connect failures attempts is exhausted. We provide operators a URL
link to removed them off the block list. This only happens with
non-full time low end operators and sites. ETRN will not work at this
point for any old mail that was already bounced.
Once upon a time, we had an auto-unblock feature where if an email
came in from a outbound blocked host, it would auto-remove the blocked
host for future outbound activity to the host.
That worked very well for good guys, except that BY FAR, the auto
failure to connect block feature was a super major success against bad
guys. So the feature to auto unblock a host when new inbound mail
negated the overwhelming worth gain in blocking bad guys with outbound
failures. So we basically made it a manual removal idea and combined
it with the URL LINK for good guys to use which sends a notification
message to the local operator to remove the host from the list.