It's a classic source/sink problem in fluid mechanics. You can't suck out a candle.
Most everything in single phase, low Reynolds number flows is about the interplay of inertia and viscosity. Viscosity acts to run out all motion while inertia is essentially your kinetic energy. If you're going to suck flow back through a nozzle it's not going to have much directionality.
Not at all. Science isn't just about observation, it's also about predictive power.
Once you run the experiment, whatever happens, happens, and you observe it and register it. But you don't want to do that. You want to come up with an answer beforehand -- I think this will happen because it lines up with what that model suggests should happen. Once you run the experiment, you know which direction it should go, but you still don't know why. Post-hoc explanations of why it rotated the way it did rob you of the chance of trying to come up with the answer yourself, so you can then find out if you were right.
I agree with pdpi's answer. But, also: given that when Feynman tried to do the experiment the apparatus exploded and flooded the lab, perhaps it's not quite so "easily performed" :-).
Most everything in single phase, low Reynolds number flows is about the interplay of inertia and viscosity. Viscosity acts to run out all motion while inertia is essentially your kinetic energy. If you're going to suck flow back through a nozzle it's not going to have much directionality.
Maybe you could exploit the Coanda effect.