OP wants code to legislate trust problems within a small, closed group, however no amount of code can ever solve trust problems within a small, closed group.
No amount of code thrown at people mismanaging their relationship with other people will ever help people more effectively manage their relationship with those people -- it just creates artificial obstacles that further exacerbate or obscure the problem.
There is simply no combination or refinement of the logic OP proposed that would ever be perfectly sufficient to avert the need for human judgement in their scenario.
What if 51% of the project simultaneously reveal themselves to be alt-right fanatics who want to take the work in a new, fundamentally evil direction, ideologically aberrant to OP? Oh that's easy - OP can just step in with his admin privileges and disable their access. And that just underscores the point -- the original voting logic was a facade, and the original trust problem remained. The presence of that logic failed to solve any problem, the reality was always a true power structure that differed from the imaginary codified structure.
Perhaps one alternative would be for OP to give up admin rights entirely - but only a fool would believe a bunch of codified rules spread across a couple of software functions could ever perfectly mediate the complexities of a community of people in every possible scenario.