I also wonder if they account for people who dual-boot Linux and Windows, and who game only on Windows or who use both depending on the game.

From my understanding, which would be made clearer if Valve actually released information on their data collection for the survey, is that the device you submit the survey on is the one that counts. So dual-booting Linux and Windows isn't accounted for. Whatever you do the survey on is the one that it records you as. The survey prompt does tell you all the data it's sending including OS. You can't modify it but you can read it.

aside from this, SteamOS and desktop Linux distros aren’t necessarily comparable enough to be throwing together in the same category. A lot of the things that make SteamOS a smart choice for the deck, where they can control and optimize for the hardware, don’t apply to desktops the same way.

Absolutely agree. It's hard to see Steam OS as a helpful metric for Linux desktop usage. The majority of Steam OS users will only use Linux on the deck and likely never even drop into desktop mode. I am curious how many people bought the dock and how many use it.