Johannes Erdfelt writes:

What's going to happen is that even if the first couple of times you fail to resolve, from that point on, as long as the queries keep coming, they will resolve. The forwarder will attempt to refresh the zone way before it expires, and unless the zone timeouts are wacky, it's a sure bet that the forwarder will hit the working nameserver again before the zone expires, and the refresh its local cache.

It's not. It's only two lines of code that must be removed in order to try every relay on the MX list. The argument for NOT removing these two lines is that if you've got a completely dead domain with a crapload of listed MXs it's going to take a loooooong time for each message to be thrown back into the queue and rescheduled, since each one of these MXs will have to be tried, and timed out.

I believe that it's far more preferable to have some messages take two or three tries to get through, then to have a constipated mail queue when the primary circuit to hotmail.com goes down. In the past I've found an amazing correspondence between proper DNS setup, and the overall competence of the same organization. Chances are that if someone takes diligent care to properly set up DNS -- with explicitly designated fallback MXs -- then they have a contigency plan for all sorts of emergencies, with any unavoidable downtimes being as brief as possible.

Although there are other things that will prevent a dead circuit to hotmail.com from plugging up the entire queue, it's still better to minimize the degree of pluggage as much as possible.