There are many time-sensitive mission applications for persistent satellite coverage, including dynamic and unpredictable events such as natural disasters, oil spills, extreme weather events, or geopolitical conflicts, which may progress rapidly and require frequently-updated information to co-ordinate the ground response. Reconfigurable satellite constellations can provide on-demand regional coverage by maneuvering orbits to focus passes over the area of interest. In contrast, traditional satellite constellations cannot maneuver to pass over specific ground locations, meaning that achieving persistent coverage spanning all possible locations of interest globally results in a requirement for thousands of satellites. This would present prohibitive costs for many applications, as well as contributing to worsening issues of space traffic management and congestion in Low Earth Orbit (LEO). Incorporating reconfigurability into constellation design allows for responsive maneuvering of satellites into repeating ground tracks (RGTs) over a location of interest, simultaneously reducing the required constellation size by improving the utilization of individual satellites and providing flexibility in the achievable ground coverage. Past work on reconfigurable constellations (ReCon) demonstrated average cost savings of 20-70% compared to iso-performance static constellations, although the complexity of the solution space for the design optimization process limited the maximum size of constellations that could be evaluated. In this thesis, a probabilistic performance metric is developed to compare constellation designs, adopting principles of reliability-based design optimization to quantify the confidence level that reconfigurable designs will outperform iso-cost static alternatives and by what margin of performance. The results show that 74.2% of reconfigurable designs outperform iso-cost static designs with a confidence level of 90% or higher, and with a margin of at least 10% improvement in the level of performance achieved. Computational intensity of the model presents the major constraint upon the size and complexity of simulation cases that may be modelled, so variance reduction techniques are applied to lower the standard error of mean performance in the output, allowing for a reduction in optimization size and runtime while maintaining the same level of error in the predicted results. Decision options for the operational phase of a reconfigurable constellation are presented and assessed to characterize how satellite operators must weigh mission priorities to evaluate trade-offs between propellant conservation and improved coverage of high-value targets.Ph.D.