Imho, it's about learning how to follow a line of enquiry to a useful conclusion.
In terms of self-led teaching, this is useful because it places the responsibility of learning on the student (i.e. it equips the student with the skills necessarily to learn independently).
If we are able to question _why_ we're doing things a certain way, we can work out whether our reasoning is watertight .. or whether it needs to be revised (because it has 'leaks').
The truly beautiful thing about this style of educating, is that the path of inquiry is dictated by necessity - that is; there's not a teacher prescribing what to learn.. the student is learning the subject matter because it's relevant to the task at hand.
I think this helps to reinforce what's being learnt, and also helps to create passionate learners.
But yes, unfortunately, it drives people nuts, unless they have a penchant for this sort of thing.
Even if it drives them nuts, it works, albeit the student tends to lose focus if s/he gets upset or frustrated and so the whole thing can take a long time.
The "Socratic method" as a teaching tool is one thing, and I actually rather like it.
The actual method Socrates uses in the Socratic dialogues is something similar, but less constructive. Basically he asks someone the meaning of "justice" or something, and then uses questions to demonstrate all the flaws in the proffered meaning of "justice". At the end, all you're left with is not a better meaning of "justice", but just the sense that what you had to begin with doesn't work.
Sometimes it's just easier to say, "Oh, you want to support local variables in your compiler? Easy, just implement your symbol table this way and support for local variables will just fall out as a side effect."
I learned this approach from my wife, incidentally, who is a teacher.
I'm a math graduate student. I was also once a philosophy student, and so I have a soft spot in my heart for the Socratic method. Unfortunately, it doesn't work so well in my teaching, because the poor snowflakes have been so well-conditioned to avoid thinking. After a year or two I gave up and became part of the problem.
It takes me about thirty seconds on average to sketch the solution to the typical algebra homework problem. It'd be more instructive to lead them through it interactively, but I just can't make the ROI high enough to make it worthwhile.
I see this a lot with students in introductory physics. If they come to me and say "I don't get it" or "I can't get started" I actually invest less time than if I get a good student who has hit a wall and can't get through. The student who won't/can't start will get sent to the tutoring center and I will spend the time with the kid who works hard. There's only so much of me to go around and I want to make it count.