I agree with everything you said. In real life, though, is ISO 25600 REALLY useable on the 1DX? I was shooting soccer Saturday night and images where I used that setting are not useable at all. In fact, I noticed on the 1DX the higher the ISO from 10,000 and above, the worse the highlights got blown out. However, I agree with your assessment regarding the 7D. THAT for sure needs improvement.
Well, I am not sure how one might define "usable". ;P You lose about 1 stop of dynamic range for every stop of higher ISO. If at ISO 100 the 1D X gets the same 11.7 stops as the 5D III...then you lose as much as 8 stops of DR by the time you ready ISO 25600 (lets say 7.5 stops, as its not always a full stop depending on exactly where the "real" ISO level falls.) If you start with 11.7 at ISO 100, then you have a mere 4.2 stops on average (and 3.7 at worst) by the time you hit ISO 25600. As far as I know, thats simply a matter of physics, and not much you can really do about it.
However, read noise at that high of an ISO setting is going to be pretty much nil (its around 2.5-3 electrons worth in most Canon cameras at ISO settings above ISO 400). At that level, its the same as Sony Exmor sensors at ISO 100 (which fall around 3e- read noise). As such, you should be able to underexpose slightly to preserve the highlights, and have the ability to recover whatever shadow detail there is without experiencing pattern noise introduced by the electronics...unless your lifting by several stops. Again, as a matter of physics, the shadows at such a high ISO setting will contain low, "shotty" detail...in that the physical nature of light will dominate, and its pretty much random hit or miss whether a pixel actually receives enough photons to even register a minimal signal or not.
It wouldn't matter if it was a Sony Exmor sensor either...at this level, your battling physics. The results will be roughly the same. The only way to really improve the performance of such high ISO settings is to radically improve quantum efficiency (Q.E.). An increase of 10% is barely going to do much, so were talking some radical improvements....backillumnated full-frame sensors, thermal cooling to -80C, and more precise manufacturing with higher-grade materials (copper wiring, silver interconnects, etc.) At 80%+ Q.E., ISO 25600 might actually become "really usable" as you put it...but you would be paying in first born children (AND the gold spun from straw that you got in return for the child), a few arms & legs, and perhaps an organ or two to own such a gem in the first place (at least, until full-frame backilluminated sensor design and thermoelectric cooling efficient enough to achieve -80C in a consumer device became "cost effective"...)
All things being equal, 4+ stops of DR at ISO 25600 seems fairly usable to me, considering that you have the option in the first place to cover your ass in the event that you actually need it. In the sample photos I've seen of ISO 3200, 6400, and even 12800 from the 1D X, the results are phenomenal, and contain far less noise than my 7D does at 1600!!