> Consumer audio can go up to a 192kHz sampling rate, which can't even record 100 kHz sounds.

That's probably because the Nyquist frequency (the maximum frequency that can theoretically be reproduced) is 96 kHZ at 192 kHz sampling, but the aliasing for that frequency is going to so terrible and unpredictable (what is the likelihood that always sample the max point in the sine wave) that you wouldn't consider anything close to the Nyquist frequency. You might get something workable for 20 kHz (~10 samples per cycle), which is the edge of human hearing, so that quiet overtones / high frequencies in percussion like cymbals show up.

If this is getting used to identify problems based on sound signature, not just "hey, the FFT shows a spike at 100 kHz", 10 samples is probably a bit low. So 1 MHz sampling rate at a minimum for 100 kHz.