Autocontrol Reverse Engineering

I have started reverse engineering the Autocontrol schematic.  It is an astonishingly simple and elegant bit of design.  It does rely on some of the less salubrious properties of Germanium transistors but it can be implemented in Silicon.

It essentially consists of a 10 megahertz LC oscillator, coupled to a frequency modulation Foster-Seely discriminator, coupled to H-bridge that drives the servo motor.  I double there is 50 components in the entire thing and a dozen of those are AC shorting capacitors.

I am having difficulty in understanding the oscillator circuit.  It is unusual in that it has a conventional biasing network for a PNP amplifier but the feedback is applied to the emitter via a lower value capacitor.  I can't quite seem to get my head around why the oscillations are at 10 MHz.  It certainly seems to work out that way in the simulator I am using (LTspice.)