My boss sent me an article which might be interesting for you.
It cites four discussions from here.
I couldn’t find the author, so he might be a silent reader.
Julius Uy, if you see this, we’d be delighted if you joined the discussions here. Your perspective would be very welcome!
The article was posted on Linkedin and Medium
IMHO the Medium article is more visually appealing, but there are no comments yet. However, the reading time on Linkedin is somehow shorter. 
This is great that this Discourse has such a good visibility. I now regularly find it in search results when I search about something Fortran related.
The comment by Katherine Zacapa below the publication on LinkedIn reads (emphasis added by me):
When 80% of climate models run on Fortran and a full migration would cost $15M plus slow things down, maybe Fortran isn’t legacy tech… it’s just the tech that quietly wins.
It makes some good points, and I can confirm that some of these are raised when decisions are made to stay with Fortran in atmospheric research (even when we’re looking at building a model ~from scratch).
What bothers me a little about a lot of calls for migration (aside from Fortran-specific myths) is that they seem to operate on the assumption that moving from having a diversity of languages to a simply having C++/Python is somehow desirable.
I think it’s good to have less languages in the project. I would also remove Python and just use C++, then you only need people with C++ expertise, not C++ and Python and Fortran.
Our goal here is to ensure that the single language to use in a large project can be Fortran. So we need to have interactivity like Python (one of LFortran’s goals) and we need to have good modern hardware support (GPUs) and some language extensions like the traits proposal will also help.
Within a project, it makes absolute sense, but the sentiment I’m encountering more and more in conversations like these is that of “let’s migrate everything (not just this one project) to the same 1-2 languages“.
