Well hell, sometimes it just feels good to step barefoot into it knee deep just to wiggle your toes and feel it moving between them. On my good foot at least..
I'm going to throw some ideas out there as food for thought.. not as gospel though some will be.
When considering the sRGB vs. Adobe98 question consider this:
It's conventional 'wisdom' that white balance also doesn't change in raw. That we can change the white balance after the capture with virtually no image degradation. I find this to be true 'most' of the time. There are exceptions and allow me to explain why I "think" this happens.
Sensors are analog devices. They capture an analog source (light) and through the process of amplifiers, D/A converters, and other electronic circuits we end up with "raw data" on our flash memory cards.
Analog devices have a design threshold where they perform their best (or capture the most data) when operating inside a certain range. Sometimes different ranges produce different sets of data. The ability to capture data on each side of this median indicates the quality of the device, and is often related to the set range of said device. Some DSLR's (all cameras, but using these as an example) capture X amount of data on each side of this median, others of higher quality (due to better design or higher quality components) capture more.
I'm sure we've all noticed a entry level DSLR such as an older Rebel XT350's raw file could only be adjusted so far on each side of this median without serious degradation, but a better quality more expensive DSLR such as the 1ds Mark II (same era) capturing the same image.. producing a like raw file.. is able to be adjusted far more to each side of the median (use exposure to visualize) before experiencing the same image degradation (think color saturation/noise). Of course a full frame vs. crop sensor enters into this, but I think the quality of the sensor and supporting electronics (amps/filters/D/A converters/etc) matter as well. Maybe more.
When I explain this to students I use the term "fat files" and "skinny files." A 1ds Mark II produces fat files compared to the skinny files of a XT350 Rebel.
Now.. concerning white balance. If we're in AWB and it gets 'close' to the right balance, we won't be seeing any image degradation from correcting WB in post with raw files. But, if the AWB gets it wrong, or you're say adjusted to 1900k and shooting in 8000k.. the difference can be more than 'some' cameras sensor/supporting electronics can make up for. I've confirmed this with tests, and suspect it when it happens by chance by looking at the exfil and other numbers. A 1ds Mark II can correct much further for WB than a XT350. Usually when the capabilities of the camera are exceeded it's limited to just one color channel.. red seems to pop up more often.. and most often its just enough, that when combined with the other two channels within spec.. it's a very small difference. Often the image just doesn't look right and you can't put your finger on why.
I think in the above case we're pushing the boundaries enough.. that it gets pretty technical to explain and would require deeper knowledge than Canon/Nikon/Etc cares to give.. and it happens so infrequently.. that they know it happens.. but choose not to acknowledge it does. So they say it doesn't matter. But if it does happen, and I think it does just by virtue of understanding the electronics involved, there is probably an engineer out there working for them who speaks out once in a while. Or tries to explain the concept to a select set of individuals.
I would guess the sRGB/Adobe98 issues are even more on the fringes.. but it would depend on the individual sensor/supporting electronics involved. Something that 99.5% of the time makes no difference.. but that .5% of the time when the scene/settings/exposure pushes the limits of design.. maybe.
Well.. it's starting to dry and cake around my feet.. so I'll move on..