[noggin]

Did some more tests and found the following:

The 4k TV streams, which activate HDR HLG correctly, are decoded using am-h254 (HW), Deinterlace-Methode: hardware

I suspect you mean am-h265 (HW)?

The 4k Youtube streams, which do not activate HDR correctly, are decoded using am-vp9 (HW), Deinterlace-Methode: hardware

Yep - h.265 hardware decode for h.265 streams, VP9 hardware decode for the VP9 streams (these are both progressive sources so the interlace report is a red herring - there is no interlace to deinterlace)

However how the h.265 and VP9 streams flag things like colour gamut (Rec 601 vs Rec 709 vs Rec 2020, EOTF (i.e. SDR vs PQ HDR10 vs HLG HDR etc.) and other stuff like mastering metadata) is likely to be handled differently in the two video standards, and thus handling of HDR content (and Rec 2020 for that matter) in the two codecs will depend on the hardware and software implementations of the h.265 and VP9 decoders and playback chains.

Just because the chain for h.265 hardware decoding and playback supports HLG, there is no guarantee that the equivalent chain for VP9 will - as they depend on different implementations (i.e. the h.265 playback implementation may be more advanced than the VP9 one) It's entirely likely that this can be handled by updated software / firmware rather than hardware (as the underlying video decode is the same - it's just a case of detecting the colour gamut and EOTF flags in the h.265 and VP9 streams and ensuring they are obeyed (if converting gamut and EOTF) and flagged in the HDMI output correctly.