Crash of CEncoder (VCE: AMF Error 4) when the monitor is connected to an unsupported GPU in a multi-GPU setup

I have been using a desktop with a supported (dual) AMD Radeon (RX580) GPU, which correctly works with ALVR when I have a monitor connected (using DisplayPort) to the AMD GPU, but **when I switched the monitor cable to the internal GPU** (Intel(R) UHD Graphics 630), which obviously isn't supported by ALVR, I started getting a **following error after launching the SteamVR server** (which also always switches to the Safe Mode, and has to be switched back manually), in both latest experimental versions 8 and 9:

`Failed to initialize CEncoder. All VideoEncoder are not available. VCE: AMF Error 4. m_amfContext->InitDX11(m_d3dRender->GetDevice()), NVENC: Failed to load nvcuda.dll. Please check if NVIDIA graphic driver is installed.`

I have two issues with this behavior, as it's a completely normal setup to connect the monitor to an iGPU (e.g. to save power, or to lessen the load on other GPUs - which improves computational performance):
1. There is literally no other report of this error anywhere, so it took me several weeks before I accidentally realized that this was the source of the issue (and not just some driver updates). I believe **this behavior should be documented in the program itself**.
2. If it's possible and easy to implement an enumeration and a selection of a GPU, it would be nice to be able to **select a GPU to be used** on a multi-GPU setup.