On September 5, 2019, the Veslemannen unstable rock slope (54,000 m³) in Romsdalen, Western Norway, failed catastrophically after 5 years of continuous monitoring. During this period, the rock slope weakened while the precursor movements increased progressively, in particular from 2017. Measured displacement prior to the failure was around 19 m in the upper parts of the instability and 4–5 m in the toe area. The pre-failure movements were usually associated with precipitation events, where peak velocities occurred 2–12 h after maximum precipitation. This indicates that the pore-water pressure in the sliding zones had a large influence on the slope stability. The sensitivity to rainfall increased greatly from spring to autumn suggesting a thermal control on the pore-water pressure. Transient modelling of temperatures suggests near permafrost conditions, and deep seasonal frost was certainly present. We propose that a frozen surface layer prevented water percolation to the sliding zone during spring snowmelt and early summer rainfalls. A transition from possible permafrost to a seasonal frost setting of the landslide body after 2000 was modelled, which may have affected the slope stability. Repeated rapid accelerations during late summers and autumns caused a total of 16 events of the red (high) hazard level and evacuation of the hazard zone. Threshold values for velocity were used in the risk management when increasing or decreasing hazard levels. The inverse velocity method was initially of little value. However, in the final phase before the failure, the inverse velocity method was useful for forecasting the time of failure. Risk communication was important for maintaining public trust in early-warning systems, and especially critical is the communication of the difference between issuing the red hazard level and predicting a landslide.