I'm with you in principle, but I think it's unlikely that Valve are building the game themselves, given that they haven't done much of that in ages.

It's reasonable to think their first priorities were finding a development studio [Edit: or even in-house developers] capable producing a good game, and helping them to do so. If the developers are most familiar with Windows tools and APIs, then the path to a successful game would be letting them use those, at least to begin with.

Let's just hope that they're being guided along to way toward design decisions that make a native port relatively easy if the game turns out to be good.

Edit:

The project is reportedly led by "IceFrog", which looked like a studio name when I first read it, but it's apparently a person. So maybe this is in-house development after all. Great! It would be nice to see Valve making significant games again.

Nevertheless, gathering a team with the talent and vision to make a good game is harder than finding people who can learn a certain API or platform, so if they have the former, it would make sense to let them target the platform they already know and get the game out the door. Doing it in-house just makes it even easier for Valve's linux folks to guide them in design choices that would simplify multiplatform support later. (Cross-platform development isn't all that hard if you plan for it from the start instead of painting yourself into a corner.) If the game is well received, it would then make sense to invest more time into training the devs on linux and doing a linux-native port.

Or to put it another way: Yes, Valve has an OS that keeps them independent from Windows, but that's just one tool in their kit. Proton is another tool. That gives Valve flexibility in how they bring a game to market, and how they prioritize/schedule various phases of the project. This still-unannounced game might be Windows-only for now, but I would not assume that will be forever.