JasperVriends you're right that Google is unclear about that, but an upper limit could be considered and tested, to see if/how it affects results. When a discussion has thousands of posts, I don't think it makes much sense to load all of them.
Another thing that might be worth investigating is to load the "answers" only when the HTTP user agent suggests that the navigator is a search engine bot, or Google directly. That would avoid loading unneeded information to users, although I can think of problems in detecting whether the client wants that information or not...
You're the expert, anyway, I don't want to teach anyone ๐