I think the *big* problem is that, for the player, it's impossible to distinguish between "Stuck because you've missed/haven't figured out what to do next" and "Stuck because the game has failed to give you adequate information on what to do next". Both are frustrating, but the first is a challenge for the player to solve, wheras the latter requires either a guide or luck.
The risk, though - and it's a trap I've fallen into - is that if you can't tell between the two of them, the average player will assume it's the latter. You need to have a lot of trust in a game to consider the possibility that the problem lies with you, and many games simply fail to earn that faith. And as soon as a player looks at a guide once, the path's open for them to keep looking, potentially ruining later puzzles even before they've given them proper consideration.
From a game design point of view... do you *assume* that a player will trust you? It's a brave move; it might produce better games, but it also produces *riskier* games. Not least because, of course, you as the designer may have judged your clues wrong, they might be much harder than intended (it's very difficult to step back and judge a problem you've created objectively, since you'll inherently already know the solution). It's a much safer option to err on the side of caution and be a bit too obvious.
