Ever since the load balancer came into play, all hosts starting sending through mailrelay.mcs.anl.gov, including our list servers. I would like owney (and lists.ci.uchicago.edu) to be able to send directly to the internet again. Going through mailrelay is adding an extra hop to an already long hop list, and it's creating a bottleneck on mailrelay. I had a complaint yesterday that an internal mail message took 10 minutes to get delivered, and it looks like it got caught in a busy mail queue on mailrelay. I've also tweaked some settings on mailrelay (and in bcfg, natch) to up the number of concurrent delivery attempts to hosts designated "relay" in /etc/postfix/transport. The default is 20. I've checked, and zimbra and owney are both configured to accept up to 50 concurrent connections from an SMTP client, so I added relay_destination_concurrency_limit to be 50. I'm hoping the combination of these two things will help. If it still seems slow, we can configure zimbra, owney, and lists.ci to exempt mailrelay from the connection rate limits and up the concurrency again.