There are many ideas in software design that are considered good practice. However, research is still needed to validate their contributions to software maintenance. This paper presents a method for examining software systems that have been actively maintained and used over the long term and are potential candidates for yielding lessons about design. The method relies on a criterion of stability and a definition of distance to flag design characteristics that have potentially contributed to long-term maintainability. It is demonstrated by application to an example of long-lived scientific software. The results from this demonstration show that the method can provide insight into the relative importance of individual elements of a set of design characteristics for the long-term evolution of software