If only nature were as cut-and-dried as simple thought experiments. Prisoners' Dilemma for instance: its likely the warden will lie, and punish you anyway. This changes the whole game. Conclusions about evolution based on mathematics are naieve to use such simple models.

I think you'll find that this is not a useful way of thinking. There is always something else you could try and account for. To save your thinking until you can account for everything simply results in no thinking. It's far better to fully understand the simplest model possible and add complication as nessisary, than to start with the hardest system. That is how science moves forward.

Sure. But the simplest model for creatures evolving must include statistical behavior. Every decision is not driving evolution - its a series over (life)time, with lots of noise in the result. E.g. If sometimes behaving selfishly works well and sometimes it doesn't, likewise altruism, the statistics will drive evolution, not the logic of the situation.

In the end the Markov chain of of event-decision-state change will favor those decisions (genes) that avoid the early-death state and favor the lots-of-reproduction states.

If I understand the article correctly the argument is that ZD strategies don't do well against each other, so they can never dominate a population. Therefore they will not exist evolutionarily. 

I don think that conclusion follows. I would expect the strategy to wax and wane around an equilibrium in which the proportion of other ZD players drops the value of the strategy to that of other good strategies. It's similar to many investing strategies that work well when a few people are using them but not when everyone is. These strategies never die, they oscillate from feast to famine.