Read these 2 things:
Why does Shotcut not use hardware accelerated video decoding?
The hardware decoding and scaling on Windows will use Microsoft Media Foundation and a Direct3D scaler and not VA-API. Because I have the native direct Windows APIs already working (through FFmpeg), and I do not want bother trying to get VA-API working on Windows. It uses VA-API on Linux and VideoToolbox on macOS. The native APIs are generally better supported and less hassle to troubleshoot. The cross-platform claims fall apart when you need to support macOS and/or hardware scaling without memory transfers.
speed up transcodings in Shotcut
Shotcut is primarily a video editor, not a limited transcoder, and it is not simply a ffmpeg command line frontend. The coming hardware decoding and scaling will only benefit preview and not used for export because scaling should be done in linear color, which requires additional GPU filters not yet available to integrate with hardware decoding and encoding. You need to take into consideration the entire pipeline and all of its possible arrangements. Shotcut will likely never have a completely hardware- and GPU-based pipeline. Attempting to do so would limit its compatibility and take a huge amount of additional work for such a small voluntary team.