3:45pm Pacific — Starting at approximately 2:15PM Pacific and for a period of approximately 20 minutes, one of the resolving name servers at the California datacenter was not responsing properly to MX lookups made by our message handling servers.  During that time, some messages (primarily outbound messages) routed through that datacenter would have received null MX responses and therefore would have generated NDR messages rather than being delivered.  The issue with that name server was resolved, and the engineering team is currently investigating why the failover to backup name servers did not work properly.