IMO, this latter usage is consistent with both HTML 4 and HTML5 semantics. I suppose Google's approach can be read as consistent with the specs, too, but either usage takes a different interpretation of "a document" and the relationships in a series of documents.
The Google article talks mainly about paginated content and mentions that it comes in different forms. The examples given in the article suggest that here we have a sequence of pages where we should or could have only one document, like in the example of the forum thread: it may consist of many pages, ordered in a linear fashion, but individual pages of the thread might be somehow incomplete. The whole thread is the document. I think that this interpretation is different from the interpretation of next/prev indicating a chronologically ordered sequence.
The article mentions also "The first page":
>>The first page only contains rel=”next” and no rel=”prev” markup.<<
"The first page" means that there is only one. Could we have more than one first page as entry point into the sequence? Like a story with different possible beginnings converging to a common end?