The first two links are about blue light affecting your eyes. Indeed there are many light inputs that affect your eyes. You have to understand that the cells in your eye are uniquely tuned to produce a chemical response to light. That's what seeing is. One also cannot discount the effects of your brain's interpretation of various light sources -- there are lots of horrific images that you can see, but the brain damage doesn't come from a reaction to the energy in the radiation, it's from your interpretation of what you're watching.
The next link suggests an increase in Alzheimer's disease from living near power lines. I am not sure how meaningful the results are; the entire document is caveats on how they selected which data points to use, the field strenghts are vastly different from wireless signals, and the frequency ranges are off by 5 orders of magnitude.
The last link is about a device designed to receive radio signals and apply electrical current to your spinal cord. If the radio waves worked by themselves, why would they implant a receiver?
You latch on to the "large number of antennas" thing, but it's not really a big deal. All the antenna array does is create interference patterns that control the direction of the signal. The antennas are excited with the desired signal with different phase delays, and then the interference results in some spots receiving 0 power (in an ideal case) and another spot receiving all the power. A friend of mine wrote a small simulation you can play with in your browser: https://apenwarr.ca/beamlab/. (The reality is that channel capacity depends on SNR, and while these do increase the signal strength at the receiver, they more importantly reduce the amount of noise that reaches the receiver. It is easier to null out received noise in a certain direction than to transmit power somewhere.)
In the past, we did this with carefully placed antenna elements that receive and re-radiate signals from another antenna element: https://en.wikipedia.org/wiki/Yagi%E2%80%93Uda_antenna (among many other designs). Now that DSP has become more powerful, it is possible to do it in software instead of hardware, and get better results. (Phased array radar have existed since the 70s. WiFi and 4G networks also use MIMO.)
I guess my summary continues to be: some sort of mechanism of causing damage will have to be proposed and tested. Maybe things on the very low end are bad (as your study [3] suggests). We know things on the high end are bad (don't get too many x-rays). But those effects roll off, and 5G is in an area where those effects probably don't exist.
Anywhere current flows through a conductor, radio waves follow. We have been using radio for 100 years, and we kind of know how this stuff works. I don't think there is any reason to worry (and many experts in medicine and RF agree with me).