Unexpected changes in game style create jarring experiences
A player's impression of a game comes from their introduction to the game, whether this is the first minutes of play, a few puzzles, or a trailer. The longer the player plays, the more connected to the game world they become. Things which break from their expectations will pull them out of the flow, creating an unpleasant experience, and often demotivating them.
Breaking expectations can happen virtually any aspect of the game. Here are some of the common ones.
Gradual changes in an aspect are generally tolerated by players, often even if there's no real game world explanation for that change. This can be used to introduce new rules to a complex game, or to slowly shift a setting.
Especially with theme, intentional contrast changes may be incorporated into a game without breaking the players expectations. For theme this often means switching between game worlds, such as a modern world and a fantasy world, say, for example, by walking through a closet.
Still, the expectation has to be there in advance. For example, if you've been playing a gritty police detective game, you wouldn't expect to be teleported to a magical world. But if from the start your case has supernatural elements, and hints towards more, it may work. It's still important to know the audience, as many expectations come before they even start playing the game.
Several games have successfully employed dual-modes, for example a story adventure combined with a series of logic puzzles. This can be challenging though, as the market for such a game may be reduced, with many players enjoying one aspect more than the other. It's quite possible to make a game with contrast that doesn't break expectations, but players don't enjoy.
At the level of an individual puzzle piece we further refine expectations with the law or least surprise. These are the specific expectations a player has while interacting with a puzzle piece, or other aspect of the world.
Assume everything in this reference is a working draft, there's prone to be some mistakes and inconsistencies. I figure it's best to publish and get feedback rather than write for years in secret. The terms will change, the structure will shift, and the bugs will be chased out. It'll take a while.