As the RF and AF gains has been mentioned, it is worthwhile to consider the signal strength of the interfering source is strong enough to overcome the natural tendency of the squelch/mute circuit to kill the audio output. That being the case, the noise being "heard" has apparently passed through the front-end, flattened the squelch detection and has probably caused the AGC to clamp its audio level.
The noise is an efficient "jammer", and there is one common thing about RF jammers, they work best when placed between the receiver antenna and the direction of interest. Hangflyer has clearly demonstrated that the "blocking" is ahead, i.e. forward of the antenna, which leads me to suspect that the "spark" transmitter is somewhere on the engine block. Too simple? Yes, but that explains why the hand-held approach is not revealing anything. The hand-held is working with its own pseudo ground/earth, whereas the comms receivers are using the same ground/earth as the spark generator. The energy in an untuned spark transmitter decreases exponentially with distance, and therefore it is possible to explain away the lack of interference in the ADF.
Marconi would have found the source by now, using just any old crystal and a cat's whisker detector!
So having eliminated the magneto, alternator/regulator, and engine block earth straps, I suspect that if the outer shield of the coax at the base of the VHF whip antenna has even a slightly dodgy ground, the input to the receiver will no longer be a low impedance load (50 ohms). The antenna becomes a high impedance untuned long wire, susceptible to any broadband noise generated locally - which in this case happens to be in front of it. The transmitter power effectively recreates a bond, providing the expected matched load.
So is the problem RPM or IAS related? Could even be both.
All idle speculation.