I have a collaborator who was writing Fortran this summer (and not even in physics!).

Oddly, it didn't seem like he was particularly locked in by a vast pile of legacy code; he just found it a good match for his mental model of the project.