You're completely right. There always is a risk of mission drift, as soon big money comes in. With my (too) short comment I wanted to give an answer how to give upvotes of a normal user a bigger weight.

I think you're going the wrong way in terms of consideration because you are conflating SP with control. That's true and should be true (because it was designed that way, good or bad) of the commodity on the blockchain – but it absolutely should not be true for the interaction of individuals in self-selected groups in a "community" (and by that, I mean a mechanically supported subset of communications on the platform intended for other members of the group). They are orthogonal issues and describe different ultimate intents. Conflating them is another part of how we got here.

The idea is such: create a SP curation pool and let the upvotes of this pool follow the opinion of the community. This only works if there is no conflict of interest between the interest of the SP-delegators and the vision of the community.

See, you've already found the big problem with this system. "This only works if there's no conflict of interest between the interest of the SP-delegator's and the vision of the community." Why does that even need to be a thing? There is always conflict of interest between one user and another, if only because their tastes are inherently different.

One of the grounding assumptions of Steemit as a social media platform is inherently invalid, and the sooner that we can get around this one point, the sooner we can get to building a social media platform that people actually want to use to purpose beyond a simple blogging base while chasing your audience elsewhere.

And that is thus: More powerful SP votes do not – not – describe something that is inherently more valuable as content. Not once we take into account that it is an individual who reads, and an individual who judges for themselves, and most of those individuals will not have the interests or resources that lead one to have a high SP account. Interests diverge. The underlying assumption of Steemit social judgment is that if more rich people like it, everyone else should like it, too. That is an invariant.

It is also inherently false.

Are there any communities on Steem working well in your opinion? Or even ome report on this? We'd like to learn from the best, setting up a Global Goals project community. At the moment I have the feeling it would be best, to start with a discord, bringing Steemians and Non-Steemians together. Second step then would be to introduce Steem and the possibility of doing project posts.

There are none.

There are communities of people who managed to maintain a community despite the complete absence of support in the machinery and by the underlying assumptions of how things are judged, but I don't think that any of the communities on the steem blockchain can really be defined as "working well" outside of those which are purely obsessed with "working the system." Maybe Steem Monsters, though the community interface portion of that group struggles to actually cohere beyond the interactions within the game.

Discord was the greatest thing to happen to the formation of communities on the steem blockchain, ever. And that is a sad state of affairs. The actual socialization does not occur on the steem blockchain. The actual business of community building does not occur on the steem blockchain. Communities cannot be built without the ability to self segregate, to gain distance from other groups, in some informational sense. Until something on the steem blockchain implements that kind of dynamic informational silo mechanism orthogonal to SP, then there is no real community system on the steem blockchain. It just doesn't exist.

People say a lot of well-deserved bad things about Google+, but one of the things that it was truly the killer app for was Communities. They were first-order objects, anyone could create wanted any time, invite anyone they wanted to join, members could post and talk on each other's posts without those outside the group having to see it – it was a fantastic implementation of tools which let you build social communities. Pity about the rest of the architecture. If we could get that going, we have something. But things like that need to be built in from the beginning; at this point, implementing the equivalent of sub-Reddits would be light years ahead of where we're at but is no closer than any other moonshot.

(A year ago, I wrote an article on the whole issue of the underlying assumption of top-down valuation being effectively authoritarian, and while it got a bit of traction back then I think there are way too many people involved with the steem blockchain who are perfectly okay with the assumption of authoritarian judgment of what is good and what is not good. Not surprisingly, I'm not. It also talks about some references to web of trust systems which I was also writing about at the time.)