Even though that is indeed a simple, quite trivial example, it does not really apply to the case of 30~50 GHz in combination with significant beamforming, because the average power density deposited into surface tissue can be much higher than with 2.4 GHz. You just can't focus 2.4 GHz onto a square inch (~6.5 cm^2) with a far-field-optimized antenna. According to [0], humans can sense an energy dose of 4J (0.2 K threshold, 5 cm^2 area, 1 cm penetration depth) in their palm due to the temperature rise. Assuming, say, 1 W focused onto the spot, that'd be just 4 seconds until it's perceptible. Measurable, harmful effects at thresholds slightly lower than what's perceptible don't see that far-fetched, once you consider the shallow sub-surface heating.
I don't think there'd be any frequency-dependent effects that don't stem from the interaction of penetration depth, triggered circulation (transporting heat away), and deposited power. Neglecting triggered circulation, it'd seem viable to just test thermally-mediated health effects from focused, 0.01~10 W @ 20~50 GHz (spot size limited due to a reasonable (0.1~0.8) numerical aperture of the beam focusing system) radiation.
I don't think there'd be any frequency-dependent effects that don't stem from the interaction of penetration depth, triggered circulation (transporting heat away), and deposited power. Neglecting triggered circulation, it'd seem viable to just test thermally-mediated health effects from focused, 0.01~10 W @ 20~50 GHz (spot size limited due to a reasonable (0.1~0.8) numerical aperture of the beam focusing system) radiation.
[0]: https://books.google.de/books?id=4dL6CAAAQBAJ Fig 3.a page 437