Think of a lightning stroke as the most broad-banded pulse of RF energy on the planet. Literally from DC beyond visible light to X-ray light and everything in between. Any small slice of the electromagnetic spectrum contains only a small percentage of the gigaWatt pulse energy, but that still adds up to a lot.
That energy will be found within the passband of any filter, just like the signal you want to hear but can't because of the lightning crash.
Wish I knew more DSP programming. Always suspected that if your DSP was listening to a "quiet" frequency as well as the frequency you want to hear, it could correlate the lightning waveform from the "quiet" (lightning pulse-only) channel and subtract it from the voice channel.
And that's as far as I have ever gotten with the idea. Kinda like a noise blanker, but in the frequency domain, not just the amplitude domain.
73