Modern Fortran is quite quite different from those punched cards back in the day, the latest revision being from 2023.
Coupled with a notebook like experience, one gets Python like experience, coupled with one of the HPC performance kings, and that is what LFortran is trying to achieve.
I saw labs that were converting their tools over to python start giving up for re-embracing fortran again a decade ago. Definitely expected this and not surprised by it.
I'm not sure how TIOBE works, but I wonder how much of the rise can be attributed to old Fortran projects getting thrown up on GitHub. Fortran is a fine language, and it has its strengths, but I don't think "native computation" is one of them. This kind of thing is rarely the deciding factor...
"Learning Fortran in 2023" https://www.youtube.com/watch?v=PvUQndB8R9s