The spec currently says about <track>:

"Furthermore, if the element's track URL identifies a WebVTT resource, and the element's kind attribute is in the chapters state, then the WebVTT file must be both a WebVTT file using chapter title text and a WebVTT file using only nested cues."

This requirement is in the HTML spec for the <track> element, but it applies only to WebVTT files. This is too exclusive. Instead, it should be extended for all files that are linked to from a <track> element with a kind=chapters.

It should be reformulated into something like:
"Furthermore, if the element's kind attribute is in the chapters state, then the element's track URL must identify a text track resource that is using only chapter title text and is using only nested cues."

Then the definitions from WebVTT for "text track resource using chapter title text" and "text track resource using only nested cues" need to be included in the HTML spec, too. This will make it possible for browsers to handle chapters the same way no matter what resource they come from.

As a consequence, the WebVTT spec would just refer to the HTML spec as being a resources format that supports these two definitions.