New Sony a7s3 does 4kp60.
http://thenewcamera.com/confirmed-sony- ... -and-more/I'm including this one because people often sight these sorts of chips being limited as being why 8k is not going be done in the past. Now that has been shown to be untrue, they go on price, and whatever little evidence limiting factor before they own it to be true and start attacking 16k or others (there's a reason I sound like a columnist).
If we skip back in time. Anybody remember the JVC HD1, my cheap backup phones perform like that today? Hardly any low light ability and pretty noisy. But the older 500 consumer SD model used a 1.3mp chip from memory.
Let's skip forwards, a few companies developed high speed low heat video technology, like Micron/Aptina. They at one stage dominated the camera phone sensor market. Latter Sony still cameras still had heat issues doing 1080p60. Sony signed a cross license agreement with Aptina and suddenly, like Aptina could, they could do lower energy higher resolution video too (and Red American sensors and Sony sensors seemed to have similar performances in ways). Toshiba did a 8k chip for a fullhd phone also. As posted before, the Toshiba technology has also been said to move to Sony (like MS did, Sony seems to buy up the place to get the dominant technology). Anyway, after the cross licensing within x years, Sony is also the dominant mobile sensor maker, and Aptina no longer is. Now Sony has 4kp120 sensor (equivalent to 8kp30/data rate).
So, if Sony has all this great low energy high speed sensor technology, including that in the cheap old Nokia 8k sensor chip, why not use it, or make it available to other cinema camera manufacturers years ago? Well, apart from marketing stratification and purchase redundancy (you pay more for better features, despite their true costs, and you pay to upgrade more general often as they release slowly new enhanced features that now don't cost to much to make, but can be sold at higher prices) there are a few technical reasons. Sony decided on aiming for native quality improvements. Bells and whistled, like true hdr circuits, can affect the overall native performance negatively. So, increased noise and reduced native dynamic range even. An old thing they did was to scan and sample the CMOS to high quality analogue to digital converters outside the pixel array. The ADC's would be better quality, and away from heat effects in the array and from introducing effects I guess in the circuits and adding more obstruction to the pixel (back in the days of front side illuminated sensors where circuited were either a chunk of missing pixel pad space and or other the pad itself. This means transferring charge across the chip inducing heat. Micron/Aptina had a better chip material for doing that, but that still goes do far. But now they use local ADC, and local processing across a back side illuminated chip. Performance of these new arrangements has greatly increased. At dome point the performance would be acceptable for more and more market segments. Eventually it may be within the 4.6k territory on an 8k camera. As long as you have low enough noise, a good quality HDR circuit could be accommodated, offsetting the smaller pixel pads.
Years ago I wanted to do HDR as a count of sampling, so evert time a pixel pad well gets full it resets and increases a counter, at the end of the integration period any remaining charge is sampled, but other versions of this I was considering was to continually sample. So, I looked at sampling smaller charge units continuously, ultimately per photon, though too many photons. So, just enough charge to flip the count circuit could be used. These days all these things I think, are actually out there. The point has been, how much error and noise is introduced by these circuits.
Now, a real issue I have been working on, is the relative size of the pixel to a photon bundle. As the pixel pad gets closer to the wavelength of light, the more it interferes with the photons clipping the boundaries introducing a form of irregularity/noise. As you know, photons can group in waves, the more similar photons together the higher your wave, like a laser. But also, these photons effect each other and slip and slide around each. As you concentrate light from a lens into a pixel, the density goes up. For those that can see in this near 8k detail realm, it looks pretty distorted crawly, maybe from this, but also out eyes color filter pattern makes thing uneven, but my eyes also have astigmatism also. So, very small pixels get more issue. However, as long as you can achieve 54db signal to noise ratio, you are averaging less than the lowest value of an 8 bit signal in noise, and if you have more than 12 stops at the same time, you have something worth shooting with professionally. So, it doesn't matter as much that they are small. Of course, we would like 102db+ snr for 16 bit, but we can keep dreaming, the previous figure should help with a reasonable image that you have to work on set more carefully with light control like the old days.
BTW: The 20-50nm pixel pads I mentioned before, I suspect probably map the light waves. I think they were counter based too and digital, either on or off (but I can't remember out of the sea of sensor technology out there). Even though it is digital cinema sensor technology, I imagine such things might be useful in internet of things applications, and micro drones.