The way I see it (maybe with badly corrected glasses) is that the wider ADC is largely redundant if the sum of the noise terms approaches/exceeds the differences between the recordable steps. This is a factor of the dynamic range. The sensor engineers understand this, and is why we get what we get.
It's great to see these sensors advance. The 1600 is getting a little long in the tooth
even if it's a youngster compared to the 8300.
I expect the usual improvements from dithering and stacking will roll forward for a while yet.