Some further hypotheses I'd like to add for
@Enrahim to consider are the
'amateur-player' hypothesis, which says that fidelity to real world beyond that which satisfies normal player expectations cannot add to the experience
'authorial-expertise' hypothesis, which says that an author of an imaginary world is de jure an expert in it (what they say is true in the fiction, just because they are the appointed person to say it)
'resisted-dichotomy' hypothesis, which says that more than one hypothesis can be true at the same time, they're not dichotomous
So when Arthur Conan Doyle says that a certain kind of snake that cannot climb in the real world, climbed down a bell pull to poison the victim, it's true that this snake can climb down bell pulls in the imaginary world of Holmes just because Doyle said so.