I'm not convinced the Wayland approach is actually better. It "feels" to have a lot more latency on my (admittedly somewhat old) laptop, which is somewhat disappointing since "better performance" was supposed to be one of the reasons for Wayland in the first place.
"Render as quickly as possible (even if imperfect)" is also a viable strategy. Some people seem to really hate it, but I've never even noticed it, much less been bothered by it. I certainly prefer the reduced latency. I'm probably not that special. It very much seems like a trade-off thing, rather than "this is the one true correct way"-thing.
Of course, it would be better if Xorg would support both, but that also applies to Wayland.
One of the unfortunate things about Wayland is every compositor will have its own quality of implementation affecting things like latency.
With XOrg, especially in the pre-compositing days, you could choose whatever WM you want and it wouldn't have any impact on the rendering performance of X clients. Once the Composite extension was added and everyone started running composited X desktops, that started to change, and the increased latency already started appearing - in an arguably worse architecture than Wayland because there were often three processes involved with lots of IPC/context switches per draw: X-Client->X-Server->X-Compositor->X-Server->CRTC. At least in Wayland it's more like Wayland-Client->Wayland-Compositor->CRTC.
If you're unhappy with the rendering latency of your Wayland sessions, it may be worth trying alternative compositors... they likely vary significantly. The Valve/Steam folks have made a minimal one specifically optimized for games/low-latency [0]. I doubt the SteamDeck would be seeing as much success as it is if Wayland were so problematic in this department.
If most WMs are slower out of the box then Wayland is de-facto slower. "But actually, Wayland is a protocol, and you can use this specialized WM optimized for speed" doesn't change much about that.
I don't use a compositing WM. I would have to check how the performance compares to a compositing WM.
It's just one more thing that's worse-out-of-the-box with Wayland that I need to find a solution for, while at the same time solving exactly zero issues for me.
"Render as quickly as possible (even if imperfect)" is also a viable strategy. Some people seem to really hate it, but I've never even noticed it, much less been bothered by it. I certainly prefer the reduced latency. I'm probably not that special. It very much seems like a trade-off thing, rather than "this is the one true correct way"-thing.
Of course, it would be better if Xorg would support both, but that also applies to Wayland.