Wikipedia's costs are extremely small relative the service usage. They don't have a lot of active development (relatively), and they don't need a ton of expensive infrastructure (relatively). Twitter's annual operating costs are in the $2+ billions. Wikipedia's are in the ~$60 millions. Not to say a non-profit twitter couldn't exist - this just feels like an unfair comparison. Just because two services are popular tech services doesn't mean they are apples to apples to operate.
We have an example of a service similar to twitter that's effectively free to run due to it's distributed nature in Mastodon/ActivityPub. Being financially sustainable without having to resort to dark patterns could just be an engineering question.
Dorsey has done three recent podcasts (two on JRE, one on Tales From the Crypto) where he states that he thinks there will be a decentralised blockchain based Twitter competitor in the future. I've taken that as a given for years, but to hear the CEO spell it out in public is pretty crazy. He wasn't even cornered into it, but rather brought the topic up himself.
I read that as Dorsey pandering to the audience. There is no room for a blockchain-based twitter competitor, because there is no sensible application of blockchain to the domain.
In 2019 you're more or less correct, but only due to cost and time constraints. Short-form public messaging on the blockchain already exists[0], it's just prohibitively expensive and slow. Imagine a blockchain with cheap, fast transactions. This should be harder to censor than other distributed solutions like a federated network or a similar scheme hosted on IPFS since you're taking the whole chain with you in the event of censorship, whereas the other schemes allow just the hosts of particular data to be targeted. Even before the economics work for full-on blockchain twitter, we might see a cheaper blockchain/IPFS hybrid where at least some metadata is censorship resistant. Unless the government wins the war on information, I'm reasonably convinced that a better blockchain twitter is coming.
I'm sure a gigantic portion of Twitter's operating costs are the once necessary to monetize their users. I have a feeling the product that is for the users would be much closer to Wikipedia-level costs.
It's not a very good metric for measuring actual usefulness to users at all though.
We need a business model for social media where incentives for companies and their users actually line up.