Nick, I liked your rocket in a bag as it somes up the situation nicely.
I modelled purely the behaviour that Jiff described. The net result is that a force acts (or more importantly must be reacted). If you visualise how such a motion can be produced then the situation becomes clearer.
Imagine a two stroke engine (running with just enough fuel to overcome friction). On the power stroke a force is generated and accelerates the crankshaft (0° - 180°), it then goes into compression decelerating the crankshaft (180° - 0°). If you look at the forces on the piston (assuming 0° is at the top of the stroke) the power stroke exerts a downward force created by the expanding gases, and the compression stroke exerts a downward force on the piston by the gases resisting compression.
Now look at only the crankshaft's motion (no piston or connecting rods) without assuming how it is created. You see a rotating mass that accelerates around half its cycle and decelerates around half its cycle, this matches the description that Jiff has presented to us.
In the piston engine the torque is created via the piston, and reacted via the piston. In Jiff's machine the torque is created at the driveshaft, momentum is changing and therefore a force must be acting. Unfortunately there Jiff's description ends. Jiff has put a bag around only half of his machine.
I am not aware of any means of creating the torque, as described by Jiff, without the reactive torque (and therefore a similar force) acting on the body of the machine. However my ignorance doesn't mean it isn't possible.
I think we agree as to the impracticality of Jiff's machine, but I still maintain that the original description does what it says.