SteveF
I'm not sure I see how the prop can become unbonded... There's excellent metal-to-metal contact through the crankshaft flange attachment bolts at the very least.
I tend to agrre, but do not have any detailed knowledge of how the spinner/blades/pitch controls etc are configured so was thinking of a mechanical path that might "open" with vibration. Certainly a defect and not a day one issue.
Also any change in the bonding path length could affect things, an intermitant engine ground strap etc, although I believe that was checked early on.
While the prop is an obviouse suspect any ~1/8 wavelength (or multiple) conductive element somewhat in line with the antenna could cause a problem if vibrations at higher RPM caused it to vibrate between bonded/unbonded state.
Cone Zone mentioned
I can't help but wonder if it is something to do with the ADF sense antenna just above the two comm antennas.
That could also do it if some element was not tightly bonded and was vibrating between bonded/unbonded.
Note that in any of these scenarios (other than prop which is rotating) the element would have to be vibrating between bonded/unbonded to produce a modulation/noise effect.
If it was just unbonfed the gain might be changed but that would be a static (unchanging) effect that would not cause noise.
It also need to be more or less in-line to produce the pronounced and symetrical directional effect.