The gist of this has been discussed elsewhere in the past and a de facto request for better support in the language has been made here:
Fortran is truly a multiparadigm language and it should see itself as such. Yet there are aspects in the language - object-oriented paradigm being one - that fall woefully short when it comes to authoring of modern libraries in Fortran particularly with the goals of efficient and convenient and safe consumption of such libraries. I really believe the situation mentioned in the original post here is an example of this, the issues highlighted by @aradi are all genuine concerns in professional scenarios and are among the reasons hindering the use of Fortran in many commercial environments with new codebases and the refactoring of old codes.
My suggestion will be for practitioners to make better and stronger proposals to address such issues at the Fortran proposals site.