Years ago I was optimistic that fixed shared library runtimes, such as the ones offered by Steam, would be enough to make native game development for Linux viable. But even with these, so many things simply break in places they should NEVER break, and can only be fixed by recompiling them (thanks glibc), so I no longer feel this way.
While my heart bleeds for open source, video games are fundamentally more art than function, so the expectation that the user (or the upstream maintainer, Steam) needs to be capable of recompiling them is simply unrealistic when considering the licensing & support required for this. Video games need a stable API, perhaps even more than most other types of software. A game built 10 years ago should continue to work when run today.
When considering all of this, along with the fact that Valve is the only large organization with the resources and incentive to fix this situation. The only real path forward Valve has here is to create their own custom desktop application API for Linux that goes beyond just a set of shared libraries and is fully integrated, meaning third-party game developers don't need get involved in the often messy inter-library politics the Linux community is accustomed too.
But if Valve really needs to go that far, why invent a whole new API for it? Why not just copy win32? That's the most popular API among game developers after all. In fact there is already a Linux runtime that not only supports win32, but actually treats desktop stability as a core priority: WINE. Why not just throw resources at that instead of making something new?
While I'm mostly speculating here, I suspect that this line of reasoning is what gave birth to Proton. It has also led to interesting situations, consider this: Elden Ring on release had/has a problem with micro-stuttering, this issue was patched in Proton very quickly, within the first week of release, but remains an issue on Windows to this day. This means that right now, Linux is the best platform to play Elden Ring on if you want the best performance and graphics. Much like video drivers are updated to fix bugs for individual games, Proton now fills this role as well, but with an even faster development cycle.
This highlights one of the key benefits of Proton: it scopes an entire Win32 runtime (wineprefix) to each game. A hasty hotfix for one game will not break another, each game needs to only pin the runtime that works, and no further tweaking is necessary. You simply cannot do this on Windows, Microsoft does not have this kind of flexibility.
A future where Proton outlives Windows, one where Proton expands the win32 API to include more features never supported by Windows, while I can't say it will happen, it certainly can happen, and I'm excited to see where this goes.