Testing involves more than just the most popular choices though. It's not about making sure that "most" players get a bug-free experience and never mind the rest...
*Sigh* Yes, I thought I made that clear in the second half of my statement, which you completely ignored. He wrote a program specifically to randomize the choices, which itself might not have been a satisfactory method since there were still problems anyway.
Since he also said that "the alternative is to do playthroughs and build a save game library with a lot of permutations which can be time consuming," I was mainly wondering if they could have used the telemetry data, which they already have, as that already built save game library, rather than having the devs do an actual play of DAO, which I agree does seem absurd to test one import function**.
** I kind of assume that in a situation like that they might just use commands to kill all mobs and other such things to go through the game as quickly as possible. On the other hand, doing "cheats" can also mess up your game. But even if they managed to whittle the typical 50-60+ hours of DAO down to just 10, that is still way too many hours taken away from actual game development time.