You might want to change between a render with and without VSYNC. This would be a common option in an emulator! The only way to do this is to change the renderer because the PRESENTVSYNC option can’t be changed (as far as I know) after a renderer is created.
Hi I’m new so take with a grain of salt, but as @TheRaven pointed out, you could get benefits from rendering in separated threads. I can think of a background that has some random things happening and you want that to render regardless of player input or NPC updated positions, which go mostly synchronous to calculate collisions.
So the chaotic scene elements would have their own renderers and could write to the window without waiting for collision calcs to be done. This would essentially boost time needed for physics.
But of course, this wouldn’t be strictly necessary as one could make separate threads for the collisions themselves and chaotic background world math itself, and then just throw everything in the same global render at the end. Maybe some edge case I can’t think off (related to the lib code itself or benchmarking?!) would turn multiple renderers desirable, but I tend to agree with @David_Olofson from organization standpoint.