Again slightly offtopic, but Stack Overflow has one of the best systems for large-scale discussions that I have seen; I'd encourage everyone to check out Meta Stackoverflow. Large-scale discussions like big RfCs are a pain point in large wikis, and Flow, while it had numerous benefits, would only have made that aspect worse. While the StackOverflow discussion model is very structured, and might not be directly relevant to a wikitext-based discussion system, I think it's worth looking at what kind of problems it faces and how it solves them.
Topic on Talk:Talk pages consultation 2019/Tools in use
From the POV of communities that believe in voting (i.e., not enwiki), Flow probably would have simplified some of those big RFCs. A specific voting module was planned.
It made tracking and reading through huge discussions less effective, though.
If there's a huge discussion. There are a lot of voters at votes like this one, and a few scattered comments, but there's almost no discussion on that page.
"would only have made that aspect worse" why? What is difference between StackOverflow.com?
I'm not sure what @Tgr would say, but I think that whether Flow would help or hinder depends very much on what you think Flow would have done. With the planned (but never built) voting module, it could have been very helpful for those wikis (such as the German-language Wikipedia) that use voting extensively. Consider Wikipedia:Meinungsbilder/Lemmata von Sportstätten, a vote happening now. More than 150 people have edited that page. Most of them have done nothing except add their signatures.
If, however, you expect 150 people to truly talk to each other, then nothing available on wiki really works.
A voting module is a solution in search of a problem, IMO; voting already works. I'm sure it could be made more convenient, but it is not fundamentally broken; large-scale discussions are. Sane large-scale discussion systems (StackOverflow and Reddit/HackerNews/Slashdot are the two I can think of) all involve some kind of per-comment voting and then presenting the discussion in such a way that the parts that have been judged negatively are hidden or sorted down.
The idea of hiding/voting down comments was discussed at the English Wikipedia during this consultation. It was extremely unpopular, at least among the long-time editors who cared to comment on the subject. The Reddit/HackerNews/Slashdot discussion style seemed to be considered an anti-pattern that was best suited to entertainment purposes. If memory serves, the main concerns were seeing how the conversation developed chronologically, and their experience that the most popular comment isn't necessarily the best one.
Do I need to point out the irony? :)