We analyze the main motivations that lead to the present need for supporting continuous software evolution, and discuss some of the reasons for change requirements. Achieving software that is both dynamically evolvable and dependable is our long-term research goal. We do not attempt here to propose a unified solution to dissolve the apparent oximoron, i.e. to reconcile these apparently conflicting goals. Rather, we enlighten different facets of the problem by distilling our experience through three research experience reports that act as case studies. We discuss the lessons learned from the state of the art and practice exemplified by our approaches and outline the directions of possible future research.