Sounds to me like the models are the reason people are at the community at all, but you've fallen into a kind of "binary dependency" on specific implementations of models.

Forking a model implementation, porting it to a new language or cleaning it up (according to someone's standard of cleanliness), and then having the two models duel for accuracy and code review aesthetics is good and correct treatment of the model. You'll get better models (which are ideas, not implementations). Even if the source code is part of the distribution, if the policy is "hands off, that's a magical thicket of code that's been polished and tweaked and the maintainer's gone", then it's effectively a binary dependency.

In particular, I'd encourage people who want to put effort in, to fork one of the EARLIEST versions of an eventually-successful model, rather than one of the latest / most accurate. Growing that implementation was probably a learning experience for the original author, and you probably need to have the same or similar learning experience before you can read the latest version properly.