I observed many of the things you describe, and I follow your reasoning. I've connected the batteries and it can charge them now (just a couple of Trojans right now).
Naturally, the wind died...
Last night I was seeing the cycle of furl, slow, unfurl, track back and forth, accelerate, and furl again. "Rinse and repeat", as you say.
If my thinking is correct, the furling responds solely to the thrust load from the blades. Hence, it will furl in different conditions, because several variables can add up to the necessary thrust. It doesn't have to be one specific wind speed, unless you keep all other conditions fixed.
So... the shorted connection forces the steepest rise in current, like saying "amps per RPM". This runs the blades at their slowest rate. In my case, they may be going fast enough that they won't stall. Which tells me that the generator is too small for the blades.
Then I re-connect the leads, so that they're charging the battery, and there is some more resistance, now. The rise in current is less steep (less "amps per RPM"), and the mill will run faster.
Looking up and seeing the mill furl, when shorted, at 50 kph means that under battery load, the furling may not begin until, say, 70 kph (44 mph).
I'm going to go back to the tests I performed on a lathe and see if I can gain some insight.