Adding VP9 parse element would be useful
At least to unpack superframe into single frames. A superframe will be stored as if it's single frame at container level but in theory the superframe can consist of multiple frames.
From our videodecoder baseclass point of view, this makes some confusion. For non-packetized decoder, the decoder can parse the superframe and then can split one incoming buffer into multiple frames then there is no issue with forward playback. In case of reverse playback, however, many buffers will be aggregated and stored in
GstAdapter so finding frame boundary is tricky.
Let's see the packetized decoder implementation (vp9dec, nvvp9dec and so on). The decoders are able to figure out the superframe but there might be multiple decoded frame per one input frame then the issue is still there. The mismatched number of input/output frame would make reverse playback not smooth. (gst-plugins-base!454 (merged) might be useful in that sense).