Fairly said.

FWIW, I'm a mostly retired coder with enough on my plate with my own projects and a fair degree of burnout, but I can read code. At a guess (unresearched) Piefed would be an easier codebase (and Python a more accessible language) to fork, but I'd probably look toward optional modules that could be added to standard Piefed at the discretion of an instance. Forking is a lot of duplicated maintenance.

At a guess (rule of thumb for developer estimates is double the number and increase the units. You could do it in 2 days means 4 weeks, it's easy to envisage something but the implementation details multiply) the community moderation code would likely be on the order of a couple thousand lines. Given good design docs (hard) a team of three might take a month to get to a stable minimum viable implementation and it could probably be maintained solo thereafter. Highly achievable.

Getting a solid but flexible (to enable experimentation) architecture in place is key and starts with a well formulated, constrained (core features only but with examples of how it is likely to expand) explanation of the goals. Then a mockup of expected behaviour. I'd be happy to offer constructive criticism.