When turning, the only rotation is about the vertical (earth-fixed) axis.
Then take it to the extreme case of diving straight down (here defined as
the longitudinal axis vertical). In that situation, all of the rotation
will be around the longitudinal axis of the aircraft - i e roll.

Remember, we're talking about Sammy's model. When diving there is no
rotation, about the longitudinal or yaw axis. (There is no aileron
input.) Same thing going straight up. The orientation of the lift
vector (positve values) would rotate the glider about the lateral
axis.


In turning flight with the fuselage level (longitudinal axis horizontal),
all of the rotation will be around the yaw axis of the aircraft.


Again, not according to the model as I understand it. If the aircraft
is sinking, the one wingtip is travelling faster than the other, and
therefore there is a difference in angle of attack, and in model, this
is accounted for as a roll. I'm waiting for the book to show up, but I
suspect that we're all taking this too literally, trying to justify it
in the real world.

As for the train, the only force of importance is the side force on
the rail. There is no friction (or better said, it is unabserved). The
point is that a leaning train looks like a turning train to a sensor
that simply measures force acting in a single direction.