But then again: IIRC, Starlink is much more focused so power per unit square might be equivalent perhaps? It needs to get to space with a good enough SNR.
The phasing of the array might also amplify the harmful effect, I'd say. But I'm very much not an RF engineer.
The starlink antenna is not very complicated and way less focused than what a modern 4G, LTE, 5G antenna use.
As an example, the latest antennas from Ericsson supports 2000 elements, which allow very narrow beamforming and target tracking (that is phase controlled movement of main beam to track individual phones).
Of course, my point is just that saying "it's non-ionizing radiation" isn't a magical rule that makes it safe. You can absolutely have dangerous or even lethal levels of non ionizing radiation too. (I think I should have used a simpler example - your microwave uses non-ionizing radiation too, yet would most definitely burn you if you put your hand inside it).
The microwave is also generating many hundreds of Watts, up to kW of radio waves, a bit like the mobile base station. In slightly different frequencies. It is still several orders of magnitude less than what Starlink emits from the antenna.
The phasing of the array might also amplify the harmful effect, I'd say. But I'm very much not an RF engineer.