It is arranged so that above a threshold RF current though the tank coil (which is proportional via its Q to the resonant voltage across the coil) the rectified voltage from the link winding exceeds the supply volts and feeds energy from the tank back into the supply rails.  This loss of energy therefore appears as a series resistance in the tank and serves to limit the current to safe value.   By monitoring the DC current fed back the degree of overload can be seen and being DC, is easy to flag alarms and things

It is a bit fiddly to set up at test - ideally if the link coil were 100% coupled to the tank coil , the turns ratio would be exactly equal to the loaded Q of the tank circuit, but coupling isn't perfect so the turns ratio is a bit lower than the working Q.

Apart from using this system on my 700 Watt  137kHz transmitter,    http://www.g4jnt.com/137tx.pdf  where it works exceedingly well indeed, is anybody else using the technique?  And would it benefit from further publication and a write up somewhere?

Andy  G4JNT