Based on pings to both Facebook.com and m.facebook.com, Facebook handles load-sharing by routing (non-mobile) traffic to different servers, and that's probably what the DNS infrastructure change they made had to do with. Because the mobile site was not using the DNS load-balancing—all those requests went to the same IP address—it didn't suffer from the problem. This is exactly the potential advantage that retailers want to leverage for their sites. By keeping the mobile sites active—and, of course, available for desktop traffic too—it's one additional level of security when the site has problems. Don't forget that when the site goes down, a lot of in-store functionality can crash right along with it. The lesson is that you want a separate IP address for your mobile site, so you're not putting all your IP eggs in one basket.
Facebook went down for about an hour on Monday, apparently due to what Facebook said was "a change to our DNS infrastructure." Like many DNS mishaps, the users who couldn't reach the site were scattered around the world, as many could get through when their local DNS lookups updated. But in a heads-up for all of retail, Facebook's mobile sites never went down.