The reason NetHack isn't already the Ideal Experimental Method Game is because once you learn what the right experiments are, you can just use them repeatedly each game; the qualitative differences between magical items are always the same, and it's just a matter of rematching label to effect for each new session.

Yes. That's why

So the game should generate different sets of fake names for each time it is run, and have some variance in the forms of clues and which NPC's give them.

isn't quite the perfect solution: you can still look up a "cookbook" set of experiments to distinguish between Potion That Works and Potion That Will Get You Killed.