I don't think it's about observing the filament directly, just that the filament basically being a spring is more susceptible to vibrations which would change the amount of light emitted.

I am totally guessing at that, but it seems reasonable.