I think one of the most cogent arguments here is that of symmetry. Someone mentioned it up thread (
@Ovinomancer maybe, or maybe it was
@Hussar). That is, if you assert that the rules, particularly the check system, represents a mechanical implementation of the 'reality' of the game world, then of course a check which fails must result in distinct, discrete, immediate failure consequences. This is because this system NECESSARILY must model all characters in the game, PCs, NPCs, monsters, everything. If the result of a monster failing a check when it attempts some action is an immediate discrete failure with consequences, then the same must be true for a failed check made by a PC, otherwise you've undermined the whole concept of mechanics bind to game world reality. While you can certainly make those consequences more or less wide ranging, in every specific case the GM would have to be able to fairly state that the consequences to a PC are exactly equivalent to those which would be suffered by an NPC/monster.
Honestly, I think simply thinking about checks and other mechanics in this way naturally leads to rulings like "the guard spotted you, he sounds the alarm!" This is another way in which the attempt to create PC/NPC rule symmetry, and 'rules as physics' is not really a strong approach.