Hi Arald, thank you for reading my article!

I'm not sure whether I understand the question completely correctly, but I don't think I would call it a bias.

The deterministic part of the state transition (depending on current state+action) is made explicit with post-decision states; I can see how this might be perceived as a sort of 'pre-positioning' before drawing the stochastic part (random environment information). However, the deterministic step is not really taken before the actual transition, it is an integral part of the step from s_t to s_t+1. Most textbooks just implicitly combine both parts into a single transition process.