As more new and existing tools continue to integrate the MRNet tree-based overlay network (TBON) infrastrucutre for scalable communication and data analysis, we continue our efforts to improve MRNet performance and scalability. In this talk we describe ongoing projects to improve the startup and runtime operation of MRNet-based tools. Namely, we overview (1) our current improvements to scalable tree startup and (2) our initial work on TBON performance modeling. This latter work is the first phase of a larger project on scalable, autonomous TBON reconfiguration.