Coseismic landslides are a major hazard associated with large earthquakes in mountainous regions. Despite growing evidence for their widespread impacts and persistence, current understanding of the evolution of landsliding over time after large earthquakes, the hazard that these landslides pose, and their role in the mountain sediment cascade remains limited. To address this, we present the first systematic multi‐temporal landslide inventory to span the full rupture area of a large continental earthquake across the pre‐, co‐ and post‐seismic periods. We focus on the 3.5 years after the 2015 Mw 7.8 Gorkha earthquake in Nepal and show that throughout this period both the number and area of mapped landslides have remained higher than on the day of the earthquake itself. We document systematic upslope and northward shifts in the density of landsliding through time. Areas where landslides have persisted tend to cluster in space, but those areas that have returned to pre‐earthquake conditions are more dispersed. Whilst both pre‐ and coseismic landslide locations tend to persist within mapped post‐earthquake inventories, a wider population of newly activated but spatially‐dispersed landslides has developed after the earthquake. This is particularly important for post‐earthquake recovery plans that are typically based on hazard assessments conducted immediately after the earthquake and thus do not consider the evolving landslide hazard. We show that recovery back to pre‐earthquake landsliding rates is fundamentally dependent on how that recovery is defined and measured. Clarity around this definition is particularly important for informing a comprehensive approach to post‐earthquake landslide hazard and risk.