We changed our DNS last night in a way that broke incoming email using subdomain addressing.
The underlying cause was using CNAME records and the fact that the email RFCs say to rewrite the domain part to the CNAME target (in this case web.messagingengine.com). This caused information loss that meant we couldn’t determine the actual recipient.
The rejection code we used means that mail would not have bounced, but would have been queued on the sending side.
We changed the DNS entries back, and queued mail appears to now be coming in correctly, however it is possible some mail has bounced from remote servers which didn’t re-check where to send the message.
This problem existed from approximately UTC: Jan 30 10:00 to Jan 31 01:00 – possibly up to an hour later at either end due to DNS caching.
edit: looks like it actually started a few hours earlier than we first thought – updated the timestamp and period values
