I agree with your suggestion that the contracting officer's determination of the most probable cost is flawless means that a fulsome cost realism analysis was done and is sufficiently documented, to include the reasons for the probable cost adjustments.
My point is the scenario only provides the most probable cost for each offeror. I assert that there are other aspects of the cost realism analysis that, while fully captured in the contracting officer's file, are nevertheless not being considered by the notional contracting officer in their award decision, based on that information being omitted from the scenario. At least that's how I read the scenario. Do they have enough information to determine the best value? I'd argue no, based on what is provided.
Assuming no variables exogenous to the scenario (e.g., a nonresponsibility decision), Offeror A would likely be best value under the first situation. The cost realism concern I would have would be with Offeror B, whose offer already constitutes the higher probable cost. Factoring in this cost realism concern (the omission of a large, critical cost element) would only serve to weaken Offeror B.
Under the second situation, I think it gets more interesting. Formation of Government Contracts, 4th Ed., discusses the downgrading of an offeror's proposal as the result of the Government's probable cost of performance being significantly higher than the offeror's estimate (p. 1530). How and whether that could or should happen here would, I imagine, depend on the specific facts.