A limitation is explicitly called out in Chuubo's, where the fact that everyone has 8 will to add to intentions and the maximum obstacle is 5 means that arguably anyone should be able to do anything if they try hard enough. It basically says that the HG can determine what a reasonable application of a skill is, or, so to speak, ‘ “Really effective at using cooking Skill to blow up the world with your mind” isn’t the most effective thing in the world to be super-effective at!’*
Actually, apparently an earlier version of Nobilis had the "Monarda Law". Instead of saying no, the HG can respond “How?” to prompt the player to come up with an in-universe way something might be plausible, suggest ambiguous consequences or risks with “You can try!,” specify explicit consequences or risks with “Yes, but…,” or elaborate positively with “Yes, and…”. (See this blog post which compares it with the rule of agreement in improv.) Certainly "how?" seems like a good response to "I blow up the world with my cooking", and saying "yes, but" doing so requires ingredients or supplies not found in your standard pantry seems reasonable. ("Yes, but" reminds me of how Savvyhead workspaces work in Apocalypse World, where you can make anything in your workspace but the MC has freedom to specify severe costs and risks if they find it appropriate.)
Going back to Dogs in the Vineyard, I think "say yes or roll the dice" works particularly well there because (a) the constraints of the game are well-defined and easy to get on the same page on, and (b) because the game is about interesting moral choices, not building power or killing supervillians. In fact, the practically only mechanic in Dogs in the Vineyard is a contest/generalized combat, so "say yes or roll the dice" involves more saying yes than it would in other systems. Observation, overcoming environmental obstacles, and such just works. But I think it does assume a certain level of reasonableness and good faith. In another part of the book, when talking about describing the setting, it says ‘When a player asks you, “is there a [whatever] here?” you should either say yes outright, or turn the question back to the group: “I dunno, does it make sense to you all that there’d be a [whatever] here?” ’ I think that could be a useful tool if someone's proposing something unexpected to you, such as a ritual or other supernatural effect. But there's still some reasonableness threshold, where if someone wants to invent cellphones (or destroy the world by cooking), they're not really playing Dogs in the Vineyard any more.
So, more generally, I think that the basic idea of "say yes or roll the dice" or the Monarda Law is good (in contrast to some of my historical issues with railroading ^_^), but it does depend on some combination of good faith and being on the same page. In a Chuubo's/Nobilis game where you might be messing with the foundation of reality anyway, "yes, but" can get you pretty far, but in Dogs in the Vineyard you're fundamentally not going to build a time machine, and in Apocalypse World the buts are going to be pretty severe.
I think there's another side to this than its normal "GM vs player" phrasing, which is player agreement. If the GM is saying "yes" a lot but the players have very different ideas of where they want the game to go, that can cause problems, ranging from PvP fights to long-term split parties to player-level unhappiness. The first one can be fine if both players are happy with the stakes, and some games explicitly suggest using the social conflict rules PvP to resolve disagreements on direction, but in some situations another player saying "no" to a player can be as bad or worse as the GM saying no, in a "stop having fun" sort of way. The turning questions back on the group thing can be useful in developing player consensus, but fundamentally if one player wants to save the town from encroaching fungus and one wants to be an agent of the fungus, there's going to be conflict. In this sort of situation, a "you can try!" approach where both get to work towards their goals even if only one can succeed in the end probably works better than something that feels like "no, you can't".
Overall, I really like the ideas of "say yes or roll the dice", returning questions to the player or the group, and "yes, but", but I think using them properly requires thought and good judgement.
*: (And honestly, in Chuubo's, clearly there should be some Auctoritas protecting the world's remaining integrity, so you'd need some sort of miraculous world-breaking cooking to destroy it. Obviously.)