[VLM]

My guess is the excitation power is broadband, which explains the unusually high power required.

Smooth white noise with a spectrum 50 KHz wide centered on the theoretical center of the resonator, then the 3rd harmonic can rebroacast a resonant tiny high Q slice of up to 150 KHz deviation FM as the microphone diaphragm wiggles with sound.

Imagine a transmitter design where you make a very powerful white noise signal, then filter out a little bit of some tuned frequency and pass that narrow sliver out the antenna. A historical example of this transmitter topology would be old fashioned spark gap transmitter generating noise from DC to daylight as they used to say, then couple it to a resonant antenna. "The Thing" is merely a much more elaborate and refined variation on the idea.

[madengr]

See my response above about the FM modulation; I don't think it will work like that; it would be AM. Assuming the resonator diaphragm is linear and small signal (WRT to the sound pressure), you won't get a radiated carrier 2nd harmonic. A cavity resonator (2nd order) would produce a 2nd harmonic in AM audio as the carrier rides that resonator response.

If the diaphragm goes non-linear (say from high sound pressure), then perhaps you'd get harmonics and maybe an f/2 response. At this point it's acting like a parametric amplifier, with your carrier frequency and pump frequency. Parametric amps need a strongly non-linear reactance, hence pumping a varactor diode with a strong signal.