Effective maintenance and evolution of complex, software intensive systems requires understanding how the system works and having tests available to verify the effects of changes. Understanding complex systems is difficult, and testability of these systems is often low due to design constraints, system complexity and longterm evolution. Thus understanding the system and adding new tests is difficult. Yet, especially in these cases, the understanding and tests are important to verify the system correctness over long-term evolution. This paper discusses synthesizing models from system traces and using these models to facilitate program understanding and test automation. Basing the models on execution traces allows generation of automated tests even for low testability systems. Generating and visualizing abstracted models facilitates program understanding, which helps in system maintenance.