It is nice to avoid all those REV reactions but the complexity of managing this under the hood in the solver is too hard to debug and does not produce any benefits when working with previously used EC models. Moreover, it is tricky to add_proteins and deserialize them when using this logic.