Thank you so much for the detailed feedback! Yeah, to be frank, "walkthrough-heavy" is, well -- it's a very kind way to put it. Though it's wonderful to hear you describe the poetry as a joy (couldn't agree more tbh, I think our writers did a fabulous job), it's just as you say: turns out it's tough to design intuitive, objective solutions for something so inherently subjective!
Per-word feedback was something that got batted around, but the initial sticking point for us was that in a puzzle where all the pieces start off "active" already and you just have to pick the right ones to "deactivate", how do you really offer positive feedback for lack of interaction -- or for that matter, negative feedback on interaction without it being a dead giveaway and taking away the feeling of actually solving something? (As I'm thinking about it again now after the haze of jam time though, I wonder if it couldn't be worth revisiting if we just took things in multi-word buckets instead of per-word, to give you *some* progressive feedback without going word-specific right away? Say, for every X words you correctly black out, you get some positive progressive feedback like the image brightening or otherwise slightly changing, but if any of those X words you blacked out were intended to be part of the solution then you don't get the progression, etc? Which tells you something's wrong, but doesn't tell you right away which ones you went wrong on. Hmm.. definitely worth exploring.)
Anyway, on another note, I think your point about labelling the target image with a Help button/calling it a "hint" system being a misnomer is right on the money -- it's more a necessary question, than an answer. Really good insight there.
Many thanks again for all the excellent suggestions raised here. You rock!!