Disaster recovery is multidimensional and requires theoretical and methodological approaches from the interdisciplinary social sciences to illustrate short- and long-term recovery dynamics that can guide more informed and equitable policy and interventions. The 2015 Nepal earthquakes have had catastrophic impacts on historically marginalized ethnic groups and Indigenous households in rural locations, arising in the immediate aftermath and unfolding for years afterward. Analyzing factors that shape household recovery patterns can help identify vulnerabilities and adaptive capacities in addition to signaling potential future changes. We pursue this goal using survey data from 400 randomly selected households in 4 communities over 2 10-week intervals at 9 months and 1.5 years after the earthquakes. Building on previous research that used non-metric multidimensional scaling ordination to identify patterns among multiple indicators of recovery (Spoon et al. 2020a), we investigate associations among these patterns of recovery, hazard exposure, and four domains of household adaptive capacity: institutional participation, livelihood diversity, connectivity, and social memory. Our results suggest: (1) social inequality, high hazard exposure, and disrupted place-based livelihoods (especially for herders, farmers, and forest harvesters on the geographic margins) had strong associations with negative recovery outcomes and displacement; (2) inaccessibility and marginality appeared to stimulate ingenuity despite challenging circumstances through mutual aid and local knowledge; (3) recoveries were non-linear, differing for households displaced from their primary home and agropastoral practice and those displaced to camps; and (4) some households experienced rapid changes while others stagnated. We contribute a temporal dataset with a random sample collected following a disaster that uses a theoretically informed quantitative methodology to explore linear and non-linear relationships among multidimensional recovery, adaptive capacity and change and provide an example of how vulnerabilities interact with adaptive capacity.