Rockfall hazards are characterized by their sudden occurrence, complex evolutionary processes, and the difficulty of identifying reliable precursory signals for early warning. Understanding their failure mechanisms and precursory characteristics remains a key scientific challenge in geohazard research. In this study, the Mogangling high-elevation rockfall, triggered by the 5 September 2022 Luding earthquake, was selected as a representative case. Based on long-term in situ observations, we developed an integrated analytical framework that combines ambient seismic noise with hydro-mechanical multi-source monitoring data to investigate the evolution and failure mechanisms of rainfall-induced rockfalls. A rockfall event detection method based on the short-term average/long-term average (STA/LTA) algorithm was established to identify rockfall events, while relative seismic velocity changes (dv/v) were extracted from ambient seismic noise interferometry to characterize temporal variations in the internal state of the rock mass. These results were further integrated with rainfall, hydrological observations, and slope stability analyses. During the monitoring period, a total of 329 rockfall events were identified and classified into three categories:single-trigger events, short-duration event clusters, and continuous rockfall sequences. Rockfall activity was strongly associated with rainfall but exhibited pronounced nonlinear, stage-dependent, and time-lagged responses. The dv/v measurements were highly sensitive to internal structural changes within the rock mass, showing significant anomalies approximately 2 h before failure and a rapid decrease of about 3 % during the collapse, indicating the transition from progressive damage accumulation to catastrophic failure. Integrated analysis of the multi-source monitoring data revealed that the rainfall-induced rockfall process can be divided into five evolutionary stages: the baseline stage, precursory fluctuation stage, instability stage, recovery stage, and re-equilibrium stage. Hydrological, geophysical, and stability-related parameters exhibited strong consistency across these stages, revealing a progressive disaster chain of rainfall infiltration–hydraulic deterioration–stability reduction–damage accumulation–failure. The results elucidate the hydro-mechanical-seismic coupling mechanisms governing rainfall-induced rockfalls and demonstrate the potential of integrating ambient seismic noise with multi-source monitoring data for precursor identification and early warning of rockfall hazards.