Can you attach a full dmesg log from boot until this issue occurs, with kernel options amdgpu.dc_log=1 and drm.debug=6? Can you also paste what 'xrandr' outputs before and after this happens? I wonder if EDID read fails for whatever reason.