Why can’t you just write server-side REST with the assumption (whether it’s actually true or not) you have no control over the client-side whatsoever? Then, whether you trust the “stability and long-term maintainability” of the client-side framework becomes moot. Pretend a new client is being written in a different framework every hour, on the hour. Pretend every user homebrewed his own private client.
What semantics are in your server-side HTML that isn’t derivable from the REST response? I can understand “rich formatting” of content like strong, emphasis, etc., but hopefully not tables or lists, eh?