As far as sub 2 is concerned a racer would have to cut the mile pace to about a 4:33 or so. Can this be done without any time disqualifing conditions?
There have been several papers published in the exercise science literature on how fast the marathon can be run. Some are based on mathematical progressions based on previous improvements. There are problems with this method.
Another author took the angle that if you could find a runner with the highest VO2max, the highest velocity at LT and the best running economy, then the record for a male would be something like 1:56. Of course finding that person is the key and intuitively a high VO2max tends to mean a poor (relative) running economy–just some observations from a little runnning data I have seen.