Running discover on a wav file doesn't categorize audio/x-wav as a container. If the wav file contains raw audio, it claims the "codec" is wav, not audio/x-raw.

This seems to have never worked, it breaks in 0.10, 1.0 and git master.

This is because decodebin2's stream-topology message doesn't declare it as a container either. And this is because it is not categorized as a demuxer. We probably need to separate the concept of "is a demuxer and we need to add a multiqueue afterwards" and "is a demuxer for stream-topology purposes". Then the topology can mark things that come from a decoder differently than what comes out of a demuxer or parser type element.