How to kill a grad student's love of research: Step 1

Now that I am in the midst of applying for faculty positions, I spend a lot of time thinking about the kind of supervisor I want to be when one day (hopefully soon!) I have my own grad students. Although grad school was hard work, and I am very happy to never have to write a dissertation again, on the whole it was a positive experience for me. My supervisor was really excited about research and good at motivating me, and as a result we worked well together and managed to be very productive. But I know a lot of people who have had such bad experiences with their supervisor that they've left academia altogether, depriving the field of truly promising scientists. So in my quest to be the world's best supervisor, I've started thinking about what I should avoid doing so that my students don't end up hating science.

There seems to be a fine balance to be struck with how often you interact with your students and how much control they have over their projects. Based on the not-so-random sample of grad students I've talked to, a supervisor going to either extreme of attentiveness will cause problems:

1a) Be an absentee supervisor
I've heard more than a few students (and ex-students) complain about how difficult it was to get help when they needed it. Particularly when students are just starting out, they need a lot of guidance from their supervisors. And when they don't get it, things can be disastrous... New students often have great ideas, but have trouble implementing them well because they don't have the background knowledge in the theoretical and practical aspects of their field yet. It's hard work for the most experienced researchers to design an experiment with minimal confounds, and yet many supervisors seem to expect their young grad students to do so without much assistance. The end result seems to be crap projects that have to be re-done over and over again, leading to frustration, despair, and a desire to do anything other than research for a living.

1b) Be a control freak supervisor
Conversely, students seem to be equally frustrated by supervisors that are always hovering over them and micromanaging everything they do. This seems to be a problem particularly for senior students who, like teenagers, are struggling to assert their independence from their academic parent. The joy of publishing one's work is pretty much lost when you feel like you had no real contribution to it or everything you did do was redone by your supervisor. In extreme circumstances, I've heard students describe themselves as slave labour or unpaid RAs. It's hard to enjoy and take pride in your research when you don't actually feel like it's your research.

The Solution?
It seems like the solution that would satisfy everyone is to have a gradient of attentiveness, with new and inexperienced students getting more supervision and more senior students being given more freedom. Not having had a lab full of people to supervise myself, is this actually a reasonable and practical thing to do? Would there be a perception of inequality and favouritism if some students got more help than others, even if it was because they needed it? How does one balance giving the student the time and attention they need while ensuring that the project remains theirs?