If I'm purchasing a game that I plan to play exclusively on linux, I still tend to check proton DB before I click buy, but it's becoming more and more of an after-thought.
The first 50 minutes of Tiny Tina's Wonderlands (Borderlands) on Linux and Review https://youtu.be/T_3sCcQksKQ
I'm running a Radeon 6800 XT undervolted at stock clocks and 980 mV (+ custom fan curve) and I don't think I could keep it cool without the undervolt. How achievable is that in Linux these days?
I'm also thinking of getting a steering wheel and getting back into race simming, are these kinds of peripherals well supported, or is it a game of chance?
As far as peripherals, sometimes it works better than windows and sometimes not great. I'd look up the specific wheel you plan to buy and see if it's supported or someone has built something to support it. For example, you've got piper (https://github.com/libratbag/piper) for Logitech devices.
Force feedback wheels work in Linux too. The "Linux" builds of games tend to have inferior or missing implementations, but running the windows binary through proton should work pretty well. The input layer in Linux is a bit fiddly; you need to set up obscure udev rules to eliminate dead zones for a steering wheel.
I actually built a force feedback wheel from scratch as a hobby project. I developed the firmware in Linux originally, but switched to Windows so that I could more easily run force feedback test programs. When I bother to set it up, I am probably also setting up VR too. VR works in Linux pretty well these days too, but all the subtle reliability problems compound. On any given day, 80% chance of game launching * 80% of wheel working * 80% chance of VR startinf means it only works 50% of the time. So I just boot into Windows, which works more like 90% of the time.
I mean for example, the back-end systems gave way after a doubling of requests (energy company). As a result, they spun up projects to improve performance, on the one side by enhancing the existing (commercial, expensive) systems, on the other to replace these existing systems with newer ones.
One thing they already achieved is to run API endpoints on lambdas, doubling performance (at least, I'm sure more can be achieved yet) and slashing costs by a factor of 60. While the lambdas are AWS proprietary, those too I'm sure use tons of open source software.