I disagree. This is kind of the conventional wisdom which I've heard repeated a number of times online, but it doesn't square with my favourite games and settings. I see it as one approach, but more fashion or trend than reflecting truth about good design.
For one thing, despite Jason's incorporation of RuneQuest into the "answer 3 questions" model, I think that this game, particularly RQIII, would fail the earnest scrutiny of 3-question askers if released today. It doesn't shoehorn PCs into any particular roles, but offers instead a vast sandbox. Most fantasy games are like this.
My favourite settings are sandboxes. Atlantis from the old Bard Games, Tekumel, Talislanta - I'm sure others could add their own personal favourites. Not so much games with a limited focus (Dark Heresy? Answers the 3 questions nicely, but I'm just not interested in the answers). And my favourite systems are generic ones.
Yes, a broad sandbox does make the GM and players think harder about what they are going to do before starting the campaign. But this lasts for about 5 minutes, before they are engaged in something they want to do, having used their creative faculties rather than being shepherded along.
I don't disagree about a focused approach being useful. It's one way to design a setting or game. It's just not the only one. If the designer wants to present a large sandbox, it needs to be excellent. The details needs to draw you in. This is difficult to write. Anyone can come up with a focus, but not everyone can write a 200-page fantasy setting that is so imaginative on every page that you want to play it. Think Artesia: the Known World. This is a traditional fantasy setting, heavily influenced by the writers RQ experience, which does not easily answer the 3 questions because it is so broad. But the writer is bloody brilliant. Works for me.