Lessons Learned From Developing and Maintaining Shared Astronomical Software Packages

Software which performs specific useful tasks for astronomers can have a very long life. If the user interface is simple enough and the code is portable, the use can spread quite widely, and if the APIs for the included self-contained subroutine library are straightforward, it will be included in other packages and demands on the developer grow. How can one person keep software useful and citeable for 30 years, and where does it go from there?