One tension I see here is sockpuppets. I’m all for allowing people to be anonymous (and potentially to have multiple anonymous identities). However, this starts to break down if people start using their multiple identities to launch Sibyl attacks and otherwise game the cred.
SourceCred has more “native robustness” to Sibyl attacks than “one-identity-one-vote” type systems. Actually, I suspect (and hope) that it will be easier to acquire a lot of cred for one identity than across two identities, because of the increasing returns to reputation that you mention. So I think rational agents will prefer to have one identity (or at least, only a few).
However, it’s also quite possible that someone running multiple identities can use these identities to game cred, e.g. having one “impartial curator” identity that just happens to always create big edges pointing towards the “eager implementer” identity.
However, this problem is not unique to one person sock puppet-ing. Even if we had ironclad one-person-one-identity rules, people could still form secret cliques and run the same attack. Therefore, the solution to this issue is not to try to control people’s identities, it’s to have a system that can detect corrupt dealing and moderate it. I’m thinking of @s_ben’s cred defenders as the agents who are empowered to search for these cliques. On a technical side, I have ideas for tools that will make it easy to see which identities are “flowing” cred disproportionally to specific other identities, so that cliques will be easier to detect.