Thanks for your reply @1script..
I used to combine (based on text length and word usage and simple text analysis) sentences that could be considered valid, with other sentences from later down in the thread replies (if those in the initial post were too short). This worked well for many many years.
This is UGC - really important and useful stuff can be only a few sentences long, other times it can be buried in 2 paragraphs of context.
There's nothing I can realistically / confidently do beyond this.
The decision is scary, but I can't justify writing complex parsing software to construct valid description text for Google (in a non-search context) anymore.
Google have the search context at runtime, so they can isolate from the total page content the most relevant part. If they don't then I guess we die off (I don't want to die off)
Cheers and thanks for taking the time to check out my situation.