With a bit of vision and just a few more features introduced into the language, Fortran can permanently reside in the Top 10 and more importantly be seen as the lingua franca of scientific and technical computing.
With the teams I work with where there are stalwarts in modern C++, Python, etc., almost everyone I have come across have been very impressed by modern Fortran whenever I have had a chance to walk through snippets of modern Fortran code with them e.g., along the likes of the example here.
The problem has always been a few rough edges (why implicit none everywhere, why default kinds of single precision, etc.), some missing features in the language (why no intrinsic string type, why no unsigned integers, etc.) - all toward some “creature comfort” and/or interoperability, and an ecosystem toward productive practice.
LFortran with @certik et al is the only vision I notice that can make a truly fundamental advance to build on the work by @certik, @milancurcic , @everythingfunctional , et al. with the Community, fpm, etc.