Steam has supported native Linux games for a decade.
The collective time spent on developing proton is probably less than a dozen high profile Linux ports.
The latest commit to proton 9.0 is three weeks ago. The latest commit to an experimental branch is 3 days ago and it is just an update of a wine version and no other commits. A lot of proton commits just update a version here and there.
The focus is certainly not on proton. It is simply very cheap to work on it, because it is highly effective. The steam deck probably cost them more software and hardware developer hours than proton.
(the Win32 and DX APIs are also much more straightforward to use than wrestling directly with X11, Wayland, Vulkan and the Linux audio API flavour of the month)
On Windows, good luck with running DirectX 6, 7, 8, and some DX9 and Direct Draw games under Windows > 8 without issues.
Direct Draw games will lag even under an i5.
A Proton like layer would also totally make sense on Windows, assuming that support for older Windows APIs is better in Proton than on Windows itself (which isn't a far fetched assumption).
But, by default, you'll get a slideshow in a game that would run screamly fast under a Pentium 3 and a Windows release from its era.
As for the missing libraries, if you can fetch some old Debian DVD images (just the first DVD) it will run fine.
it is not full Proton, but DXVK is used on Windows as it has sometimes better compatibility and/or performance for older titles.
Soon Wine will have Wayland support and 99% of those ports that no one is willing to update will remain stuck on X11.
Wine's version of the Windows API isn't conceptually different from any other cross-platform technology here. You could make a similar purity argument against using the Unity engine since most of its games end up being primarily on Windows too.
Proton has caused a significant increase in Linux playable games. The side effect is that it effectively killed porting companies.
edit: also all porting companies were effectively using their proprietary equivalent of proton (although often inferior) .
Until there is a Linus figure that coordinates the userspace and organises a common platform API with long term support for closed-source software, Proton is the only pragmatic choice.
Valve want to get off Windows ASAP, not necessarily waste money chasing windmills driven by silly ideology that native is better.
I love Linux, I have used it for 25 years, and even I accept that native games run WORSE than their Proton counterpart.
systemd, GNOME, mesa etc all have developers who are being paid by companies for their work (Red Hat, Microsoft, Canonical, SUSE etc). That said, you're not wrong on the 'no overarching vision' part, see Wayland.
> Linus figure that coordinates the userspace and organises a common platform API
Flatpak with the freedesktop runtimes are just this, that said some companies (e.g. Canonical) are trying to sabotage these efforts and Ubuntu not shipping with Flatpak is the biggest hurdle.
What we need is something that groups Qt/GTK, pipewire, part of systemd, part of flatpak, part of Wayland into a single library, a bit like Win32 is. And the guarantee that it remains stable even for closed source projects. For example Linux is free to change its internals and requires everything to be open, so drivers can be adapted whenever the APIs change. This is not good enough for a desktop API.