@Bigslap the clicks part makes sense to me, because short pays twice there. The tight answer paragraph that wins you the click is the same one a longer guide would have opened with anyway, so the short version isn't really missing anything the searcher came for... it just skips the padding they'd scroll past. Comparison queries are my one exception. Even then it's the structure carrying it, not the count.
@Chunkulus yes, and I think it's why the debate never settles. Averaging ranked results can't separate cause from survivor. The clean test would be the same URL before and after an edit, which almost nobody publishes because they're one-offs (I've never seen a proper set of them anyway). Your variations point at least gives me a working rule... expand when there's a real sub-question left to answer, trim when there's not.