This paper gives insights into the DynaSoS project, which aims to propose a cross-domain roadmap for systems engineering research into trustworthy dynamic systems-of-systems with autonomous components. The project follows a methodology that combines interviews and workshops with experts from various domains and a literature review. In this paper, we motivate the project, discuss five application areas with their drivers and visions, and draw initial conclusions with respect to challenges that a research roadmap should address. Our goal is to share initial results with the research community about the DynaSoS project and invite them to contribute to the derivation of the research roadmap.