I'd frame it this way: a good senior researcher knows when to give a junior researcher the question or not. First-year grad student? Pair them off with a postdoc for fast iterating on an established project. Fourth-year grad student? Push them by letting them flounder a bit and learn to find their own questions.

I agree that in practice, this simply varies a lot by discipline and advisor.