He has the integration time set to 1/8000 s.

I assume the idea is that you're effectively randomly sampling (with each shutter press) that amount of time of the sine wave. You should see a distribution of brightness if there's a signal present, unless you get unlucky and sample at the same brightness (many would be required). Capture many with and without audio signal present, compare std deviation. If they're the same, no signal.

But yes, truly terrible lab report. ;) Who knows if this data is meaningful.

I wonder if the original rumor came from a bulb with a failing ballast, with a "less stable" arc, or maybe with a mic/diaphragm integrated into it, that modulated the light by sinking some current? I always assume these more wizardly things are based in truth.