AlainCo
Currently I'm looking more at series of peaks spaced by frequencies in the order of kHz on relatively narrow ranges rather than doing measurements spanning very wide ranges. Violante et al. performed measurements up to 83 GHz in a way that makes looking at the small details difficult.
In any case, I tried using sodium bicarbonate (NaHCO3) as the electrolyte and I obtained about the same results, with peaks having spacing again of about 127.5 kHz. I'm not sure if this is theoretically expected from Holmlid, since his results were with potassium. But again, in completely different experiments.
NaHCO3 gave a strongly yellow light when the reaction was steady, by the way.
Most of the signal appears when electrolysis at medium voltage is performed (and the DC boost converter is already under load), but as voltage is increased (possibly giving slight plasma formation or micro-discharge) or a plasma reaction occurs, then the 127.5 kHz peaks appear as well. It does not seem to be related with the switching frequency of the boost converter as it remains fixed regardless of all the different audible noises the converter's circuitry makes under varying loads and voltage.
EDIT: for what it's worth, I tried checking one of the "bumps" (of those spaced by roughly 50 MHz) at about 1296 MHz and one around 327 MHz, but I couldn't find the same small feature there.
I also tried using a very sharp pencil (graphite+clays containing Si, Al, Fe) as the cathode, which made the plasma reaction easier to produce, but it did not bring any significant change other that it seemed more energetic.