Hey, one of the Enzyme co-authors here. We are already doing a good bit with Fortran and have made it a point to preserve the use of Fortran in papers. But I fully agree that more usage of Enzyme with Fortran poses some interesting problems, as well as possibilities.
To be completely frank, the one thing making it quite difficult, if not to say an art form, to use Fortran at the moment is the state of the LLVM Fortran stack. If you want to have a look at how it manifests in practice please feel free to have a look at the Makefile at one of the more recent benchmarks.
There also exists quite a bit of stuff with Fortran which is still in the works and will probably come through the publication pipeline in the next few months or so…