Not to disagree with any points in the article, but i would point out that the sciences also have cases of very old code being maintained and used in production

Not to disagree with any points in the article, but i would point out that the sciences also have cases of very old code being maintained and used in production successfully. For example we still use a kinematics code written in fortran over half a century ago. In practice parts of it get reimplemented in newer projects, but the original still sees use.