It turns out to be a class C Armstrong oscillator, and Adrian has a nice explanation of its operation on his website. To quote:
Circuit operation is as follows. When first turned on the impulse current from the tank circuit B+ conducting through the vacuum tube causes a ringing oscillation (ping) in the primary circuit LpCp. This oscillation couples to the secondary circuit LsCs which is further coupled by the pickup coil to the grid bias leakage circuit. When the phasing of the pickup coil is the correct way round for positive feedback, the grid bias leakage capacitor C1 becomes negatively charged during the positive half cycle in the primary circuit, pushing the grid voltage down and progressively restricting conduction in the vacuum tubes towards the off state with much reduced plate current. The negatively charged grid leakage capacitor C1 then discharges through the grid leakage resistor R4. As this happens the grid voltage on the vacuum tube starts to rise progressively towards 0 volts turning on the vacuum tubes with an increasing plate current. The plate current through the primary circuit LpCp again is coupled to the secondary LsCs and the cycle repeats. With the grid bias leakage circuit correctly adjusted the VTG will oscillate with a linear sinusoidal output optimised for maximum plate voltage and current swing, (maximum power transfer at the resonant frequency of the secondary), whilst keeping the grid bias currents within the maximum ratings for the vacuum tubes used.
So now it has become quite clear, but it may still be of interest to learn what kind of signal is at the grid of the tube in this setup?
If we could connect an oscilloscope that would withstand the conditions that exist there, what would it show?
Thanks in advance.