Alan, to your reply #9,
I think Paul pretty well described it with his comments on USAF g-meter observations.
For further definition, I strongly suspect that maneuver g-loads vary in sharp corners. (Duhh...) They increase rapidly as input is applied to start the rotation, which begins almost but not quite immediately. Then, at a specific point, determined by practice, input is 'feathered off' to place the model's attitude and position at exit of the turn.
Since a really sharp corner may not last 1/3 of a second, all this happens very quickly indeed.
The peak g-loading does not persist as a steady condition as far as our reflexes and observation could recognize. That peak load is what I referred to as a 'flash' loading.
As Igor, and perhaps another one or two, have recorded in high-rate stop motion imagery, the model may actually fly the equivalent of a race car's "4-wheel drift." I.e., it skids wide of the apparent radius. It all happens so quickly that the impression is of a sharp, radiused turn which the model tracks through. Some 'drift' outward after the model reaches the exit ATTITUDE is also almost impossible to detect with the human Eyeball, 2 each, Mark 1 (w/optional supplemental lens systems).
The images I mentioned suggest that the model 'points' at about 90° to its motion in the 'drift', and that its wing and tail surfaces may act as a 'braking chute' for some infinitesimal split second, until residual kinetic energy and prop thrust carry the model out the exit track...