While some of today's cameras do a good job at higher ISO's, extremely high ISO's don't produce great results. That being said, getting the shot, regardless of the quality, is preferable to not getting the shot. If you see Bigfoot in the evening in a very dense forest, would you rather a grainy shot or no shot at all? As amateur photographers and even pros who shoot life events or photo art, the quality of the shot is very important to us. But If you are a journalist, or a scientist, just getting the shot under near impossible conditions often comes first.

From the tests I've seen, ISO 3,280,000 looks pretty awful. There's very little color, lots of noise, and an extremely limited tonal range.

Whether such high ISO images are usable depends upon your application. Spy work? Probably. Portraits? Extremely unlikely to hell no! Photojournalism? Only in dire circumstances. Some reviewers say they would tolerate ISO 102,400 for photojournalism, but I think that is pushing the envelope unrealistically.

I take these camera sensitivity claims with a grain of salt. Image technical quality is usually best at base ISO. As you raise the ISO, the dynamic range is lowered. Color reproduction is muted. Noise intrudes. Latitude for post-processing adjustment is narrowed until it goes away. There is a point at which every stop of higher ISO takes away a full stop of dynamic range that could have been printed on paper from an image made at a lower ISO. That's the point of zero latitude as far as I'm concerned. I don't want to go that low.

My experience with 16 to 20 MP sensors is that I find my personal quality degradation tolerance stops at:

ISO 6400 for Micro 4/3
ISO 12,800 for APS-C/DX
ISO 25,600 for Full Frame

I try to limit this to

ISO 3200 for Micro 4/3
ISO 6400 for APS-C/DX
ISO 12,800 for Full Frame

... and a full stop below those when possible! I like to start thinking about auxiliary lighting when ISO gets above 800 on Micro 4/3, 1600 on APS-C/DX, and 3200 on full frame. I don't necessarily use it, and can't always add it, but I think about it.

These personal guidelines are based upon reviewing tests of Nikon D5, Nikon D500, and Panasonic Lumix DC-G9 and DC-GH5 cameras, including prints and dxomark.com scores. The dxomark scores correlate quite well to my own observations.

My personal guidelines may not relate at all to the D6, because I've only seen promotional/review images from that camera. HOWEVER, the sensor is so similar to that in the D5 that I doubt the high ISO performance will be very different. Most of the improvements in the D6 seem to be related to the viewfinder, AF, and other usability features. It will take a jump to newer technology to make any huge strides in sensor performance at this point.

Your observations are the same as my more limited conclusions. And I can assure you that Physics is not going to permit much more improvement in sensor technology.

The rule is the bigger the area of the sensor cell or pixel the better the signal to noise. Hence, with the same number of pixels a full frame sensor will be a full stop better than a DX sensor. Likewise a 20 MP sensor will be one stop better than a 40 MP sensor.

Nikon seems to be consistent in stating the max ISO before going to the emergency Hi settings as the highest ISO before color noise dominates. I find two stops before that as my limit of niceness.

I believe that given the nature of the IC foundry business model the state of the art sensors will be built by Samsung or Sony foundries and will be of similar design (no secrets in that business). Current sensors are well enough designed to be limited by the quantum efficiency (The percentage of photons that land on the sensor that are counted.)of the sensor cell. Currently the QE is greater than 50% only one stop bellow max possible. All of this is at near room temperature thermal noise. Liquid nitrogen anyone?

That only leaves clever post processing either in camera or computer tomake the picture look more like you expect.