The problem as I see it is that maintaining two API languages implies that one must always “chase” the other if the primary language is refactored. In a single-language codebase, a single refactor can fix all of the related code more easily, unless the maintainer is an expert in both.