When working with the static linking, it became obvious that we could actually compile most native files (almost) immediately upon starting the build, awaiting only native source code generation.

Splitting up this monolithic phase would mean that make would be given much more opportunities to parallelize the build, which would reasonably lead to lower wall clock build times.

Splitting this up would also mean an opportunity to handle dynamic and static linking more symmetrically, since we would have a compilation step (which both kinds of linking share), and then one (or more) linking step, depending on the required type of libraries.