Okay I worked it out.
I drew a few diagrams of lift/weight and the tail in various configurations and I reckon this is how it works -
The CG has to be fairly close to the CP, but far enough ahead so that with a high AoA the CP doesn't (shouldn't) go behind the CG. The tail would have to have zero or a very low angle of incidence and the wing a pretty good one, say, 4° or so.
That's the only way I can see the system working even close to being dynamically stable.