EzekielRaiden
Follower of the Way
I know I already replied, but this required a more focused response rather than my general one. Inconsistency would actually be much worse than random. Dramatically so. My "meddling Professor" analogy is also a thing where an unknown inconsistency is worse than pure randomness.For example, I think your concerns about GM denying players the possibility of learning must rely on an apprehension that the GM will make their determinations with negative bias or intent to disrupt, seeing as inconsistency alone would surely be no worse than random.
If something is random, you can learn how to prepare for randomness. Especially the kind of randomness usually used in TTRPGs, since there are tools available to the player to influence the probabilities. There will still be times where chance simply lies against you. That's part of life, and part of gaming. Statistics is literally the science of accounting for randomness so that we can still learn from it.
If something is genuinely both inconsistent and secretive, you can't learn to prepare for it. There is no pattern, by the definition of inconsistency, but players cannot know how or why the inconsistency occurs. Certainly it is possible for the inconsistency to be partially occluded by chance, e.g. the black box decrees something just secretly fails, and it just so happens that the roll would have failed anyway. That sometimes chance will coincide with the inconsistent input does not mean the inconsistency is irrelevant.
Indeed, I would argue the reverse of what you did here: inconsistency would surely be no better than random--and likely much worse. Try to "learn" from inconsistency, and you are essentially guaranteed to develop false beliefs, because (in the relevant sense) there are no correct beliefs about it. That follows directly from the fact that it is inconsistent.