[exim] using self=send causes 127.0.0.1 Connection refused

For about a decade I have been using a configuration which routes email
submitted on port 587 to port 24 on localhost, on which exim is
listening as well. It then will be sent out.

I do this using the following configuration:

smarthost:
debug_print = "R: smarthost for $local_part@$domain"
driver = manualroute
domains = ! +local_domains
transport = remote_smtp_smarthost
route_list = * DCsmarthost byname
host_find_failed = defer
same_domain_copy_routing = yes
no_more
self = send


This has been working great for many years. However the last couple of
years I tend to get the following behaviour every now and then or very
often, depending:

2018-05-11 11:59:20 2qHDGC-0004TT-26 <= example@???
H=host.example.org [192.0.2.1] P=esmtpsa
X=TLS1.2:DHE_RSA_AES_256_CBC_SHA1:256 A=login_saslauthd_server:example
S=786 id=3a16b96b-172c-36d6-c06c-04aa4b1e4d6e@???
2018-05-11 11:59:20 2qHDGC-0004TT-26 127.0.0.1 [127.0.0.1] Connection
reset by peer
2018-05-11 11:59:20 2qHDGC-0004TT-26 == control@??? R=smarthost
T=remote_smtp_smarthost defer (104): Connection reset by peer
2018-05-11 12:07:11 Start queue run: pid=17411
2018-05-11 12:07:12 2qHDGC-0004TT-26 == control@??? R=smarthost
T=remote_smtp_smarthost defer (-53): retry time not reached for any host
2018-05-11 12:07:12 End queue run: pid=17411
2018-05-11 12:24:22 SMTP command timeout on connection from localhost
(example_server) [127.0.0.1]

(..)

2018-05-11 12:52:55 2qHDGC-0004TT-26 127.0.0.1 [127.0.0.1] Connection
refused
2018-05-11 12:52:55 2qHDGC-0004TT-26 == control@??? R=smarthost
T=remote_smtp_smarthost defer (111): Connection refused

This can go on for a while before it eventually gets sent out.

It's never been a big problem since the email gets send out eventually,
after a few hours or half a day.

My question is how can I configure exim to stop refusing and deferring
the connection to itself at 127.0.0.1 port 24 and send it out right
away? Removing the retry file and restarting exim did not work.

Note that exim is receiving incoming email from the spamfilter on
localhost port 24 as well, after it's properly scanned by the
spamfilter. Remote MTA submissions and/or connection attempts on port 25
where the spamfilter is listening can cause connection attempts to exim
on port 24 as well, so at times the volume of connection attempts can be
larger than normal.

I suspect exim may just be throttling connection attempts on this port
due to the possibly larger volume of connections and starts deferring
any excess. I think I would have to increase the threshold when it
starts doing this? If that is possible.

Thanks,
Jeroen

--
Earthquake Magnitude: 4.7
Date: 2018-05-11 15:52:01.790 UTC
Date Local: 2018-05-11 08:52:01 PDT
Location: 40km SSE of Bobon, Philippines
Latitude: 6.5398; Longitude: 126.4951
Depth: 78.86 km | e-quake.org