I’m not saying it’s the first thing in the universe to happen. It is entirely possible to know from ones knowledge of physics and the world that a disk tossed in the air will fall on either side with equal probability. If you want to say then that the coin toss is part of a much wider ensemble of observations, that’s fine, but so is waking up in one of two buildings. One can compute 50% probability of being in either building on the basis of prior ensembles of observations where one had a lack of complete information about a situation - just like the coin toss. I don’t understand how this building situation is any more one-off than any such situation where one has partial knowledge. If you wanted an ensemble, you could conduct an experiment where you put people to sleep and put them into either building then see how often they guess right about their location. But no need to do it. We already know they’ll be right half the time.