Where I'm having trouble is seeing why there is any difference between
sinking, level, and climbing. From the point of view of raf, these
should all be same. I choose to set my longitudinal reference based on
the fuselage, not the horizon. If I do this, the aircraft always has a
sink rate, not relative to the ground, but relative to the projection
of the fuselage centerline at a given time (T=0) and proportional to
the angle of attack -- that is, the rate at which the aircraft "falls"
away from this projection (dT). Since it shouldn't matter which
reference frame I use to make my observations, my confusion arises
with the suggestion that the pseudo-rolling moment reverses beetween
sink and climb.

As for the approach, it remains interesting. To help my understanding,
I've been using a train. Imagine a sensor on a rail that only measures
side force. A train going straight on level ground registers zero side
force. As the rail bends though, the sensor would measure a side force
proportional to the train's acceleration. However, a straight rail
with a side pitch would register a force as well. When viewed this
way, a train rolling on a straight rail with several degrees of side
inclination could be said to be "turning." Of course, it isn't. Unlike
a curved rail, no additional power is needed to maintain speed. (Note
the limitations of my reference frame. I only see side force on the
rail, not total force.) It's not exactly analogous, but it's a step in
the right direction.