Your point about Metal 2.1 in your earlier post is interesting. Though I am skeptical that just the use of 2.1 vs. 3 is responsible for reduced performance on the Mac vs. Windows. There might be some low-hanging fruits/code changes that could help with these issues. Note also that when BG3 development started Metal 3 was probably not out at all...
Probably. Looking for info seems that Metal 3 is available even for Intel iGPUs. Put this only as reference no need to watch:
https://developer.apple.com/videos/play/wwdc2022/10066/But in the same video they insist much on multithreading. Then that reminds me that BG3 on PC is a DX11 game, a managed high-level API which multithreading is much less important than for low-level APIs (like Metal), can look for DX11 vs 12 multithread comparisons to see that on DX11 the 80% of the render is executed on the main thread anyway. So BG3 targeting DX11 probably is a single-threaded rendering title, or not optimized in that area as it takes few advantage of it on PC.
Then on Mac we probably get what Apple calls CPU-GPU overlap, this is one waiting the another one. (Again put here this only for reference no need to read).
https://developer.apple.com/documentation/xcode/analyzing-the-performance-of-your-metal-app/Solving this requires some specific work for Mac, so I am afraid we will not have it. But at least we could have MetalFX Upscaling, spatial replacing the FSR 1, and temporal to get finally that option, which at the same time get better results than FSR 2 due to the use of AI (FSR 2 is just a shader). They both requires the same data, motion vectors, so implementation should not be so hard I think.