It's worthwhile to note that there are currently driverless metro lines in the world. We do seem to be moving in that direction for automation. https://en.wikipedia.org/wiki/List_of_automated_train_system... As always, the future is already here - it's just not evenly distributed.
In particular I'm thinking about some of the procedures aboard nuclear submarines where automated systems were tried and eventually rolled back, because the automation would be fine 99.9% of the time but when it failed it would cause disaster at computer speeds instead of just at human speeds. I can definitely imagine a bug in a DAO being completely unacceptable in some domains even if it is more efficient than doing the same job with humans. (For example, in national voting)
Finally even for those cases where automation is desired and could be done by some autonomous entity, I'm not sure why you would specifically need a Distributed AO instead of just regular cronjobs on a server somewhere. Any real-world system is going to need regular updates anyway, so you end up centralizing trust in whoever can update the code for the (D)AO.