Purpose: Improvement science (IS) has become a popular approach to organizing school-university partnerships because of IS's potential to increase schools' capacity for sustainable improvement. However, little research has directly examined whether and how specific elements of IS support school improvement, particularly during and post-COVID-19 when improvement was particularly challenging. Design/methodology/approach: We draw on a longitudinal case study of a school-university partnership supporting a group of schools using IS to guide school improvement with data collected in Fall 2019-Spring 2022 including interviews and meeting observations. We compare how educators engaged with three IS elements: plan-do-study-act (PDSA) continuous improvement (CI) cycles, networked learning and driver diagrams. We qualitatively examine participants' perspectives of these elements through the lens of contingency theory, analyzing which elements were more or less successful at empowering schools to continue their improvement efforts throughout the pandemic. Findings: IS processes are varied in their resilience to complexity. Schools mostly abandoned some elements during tumultuous periods (PDSA cycles) while others were successfully adapted to sustain improvement work (driver diagrams). Findings also discuss the perceived impact of university partners in school improvement work, primarily as coaches. Originality/value: These findings are uniquely positioned to examine whether and how IS elements enabled sustained school improvement amidst the complexities generated by COVID-19. By focusing on strengths and limitations of three common elements, we offer valuable guidance to school-university partnerships about the conditions under which these elements might support sustained school improvement and how these elements might need to be adapted.