The double spending phenomenon might be rare and depends upon the consensus algorithm being implemented but I guess it's still the most important aspect which cannot be ignored or taken for granted as it can collapse the entire system if not dealt with. A lot of research is being done on consensus algorithms to minimize the probability of double spends to bring the theoretical-probability closer to zero and a smooth fallback mechanism in case of a successful double spend attack.
For reference, you could read this brainstorming by Dan and Jaekwon from Cosmos regarding DPoS algorithm that we already consider to be a pretty great.
If the system is really secure, it's hard and slow to use – so it's not going to get a lot of users.
Agreed, but as you mentioned - Dan claims that EOS will change things in terms of processing and scalability to the next level.
Yes, this is true. But I still think that pretty much all real blockchain developers understand this and try to make their best efforts to mitigate risks. It's not something that they might forget to include in their project, unlike all other things that I listed.
That is great conversation, I've read that earlier for several times to understand DPOS better.
Right! You summed it up pretty accurately! And yes that's a great conversation. We need more discussions and participation like that.
How's your proposal for focusing more on communities coming up?
I might publish a post on that tomorrow.
That's great! Eager to read it :D