Per w3c/publ-tests#55, the manifest processing algorithm doesn't seem to match the TOC requirements.

But I'm not even sure how we can fix this, as how can a machine detect the presence of a "supplementary resource" when they aren't marked? You can't just scan the resource list for non-audio content, as there's no way to infer intent (i.e., is the resource meaningful to the content or is it just some extra bit added to the PEP). It also can't be determined if text of the PEP contains supplementary content or not.

So should we just remove the rule and not try to determine if/when a TOC should be present? Leave it for real-world implementations to determine when and how they want to enforce this?