I'm using `OnAcceleratedPaint()` on Windows - this requires that the processes are all running on the same GPU; while the MSDN documentation says the handles can be used for 'multiple devices', this means 'multiple instances of ID3D11Device, referring to the same physical device' except for edge cases like SLI.
My app must run on a specific GPU, identified by LUID by a VR runtime - for physically connected VR headsets, this will be the GPU that the VR headsets is physically connected to. CEF needs to run on the same GPU.
The 'performance' vs 'efficient' preference settings available in the registry or windows are insufficient in some multi-GPU settings, e.g. when there are 2 discrete GPUs *and* an integrated GPU.