It's one thing to make a scenario more difficult without metaknowledge, but to completely block progress is bad design. It's why I consider AoD more of a CYOA than an RPG.
So, a couple thoughts.
(1) I think there is a really bad tendency to characterize what I would call "build-service" content as "meta-knowledge-requiring" content. We run into this a lot in designing Fallen Gods, and we call it designing for the Siberian -- basically, there are scenarios where you realize, huh, if the player had X skill, Y item, and Z follower, shouldn't he be able to do something ridiculous here? (One example: you come across a field of dead birds; there is an item that lets you understand bird-talk, a skill that lets you talk to the dead, and a bird familiar you can have, and you can see why conceivably you might want the three to interact there.) Often, this is very obscure, such that the likelihood of (a) having the necessary load-out and (b) coming across the scenario is so slim that the only person likely to encounter it is someone stuck by himself in a Siberian listening post with nothing to do but play Fallen Gods (hence, "the Siberian").
We view this as "build-service," by which I mean, it's in there not because the player is expected to get it, or to strategically play to maximize his chances of getting it, but because if you should happen to have that build, it's fun to have it actually
do something in that scenario.
What is so cool about AOD is that there are so many different builds that get "serviced" in that way throughout the game, and the service is way beyond what we have in FG (which amounts to a node of text and a little bonus, not an extra dungeon or something). But because players are used to the notion of 100% completion and
player-ego-service -- that all things are possible to the superman, and thus if there is an item behind a door, there has to be some way for your guy to open that door, no matter whether he's a diplomatist, a berserk, or a hacker -- AOD's build-service ends up being recast as a design expectation that you're
supposed to be getting every power cylinder, every possible XP and relic, etc., etc.
I think it is fair to say that
Vault Dweller failed as a designer as to those unhappy players because he wasn't able to teach them how to view the game in a light where it became fun rather than frustrating. But the core concept of a highly reactive, build-service-oriented game is not a bad one IMO. It's a great one. It's why, despite being someone who hates min-max gameplay, I loved AOD. In that regard, VD was wildly successful as a designer.
(2) That said, there is one feature of AOD that is very, very weird, which is that the "correct" way to play appears to be skill stockpiling. Like, in some ways, the way the game works is not that you accumulate XP for the purpose of building your character as you go, but that you accumulate XP as an obstacle-passing currency, and a side-bonus of spending that currency to pass an obstacle is that future obstacles will be cheaper to pass using the same method. It's an upside down way of playing RPGs. But the presentation of XP in the game doesn't look upside down at all. It looks just like Fallout or a normal RPG where there is no reason not to immediately dump your skill points into the skills that you want to use on some
non-specific obstacle in the future, instead of hoarding them until you know the specific obstacle you'll face. I'm pretty sure if the game had redefined the terminology, particularly for the non-combat skills, it might have been able to avoid seeming like it was broken. I'm not sure the concept is necessarily wrong (though I will say, I liked it much less than normal RPG character building), but it exacerbated people's reaction that the game was relying on meta-knowledge.