Ice cliffs are common on debris-covered glaciers and have relatively high melt rates due to their direct exposure to incoming radiation. Previous studies have shown that their number and relative area can change considerably from year to year, but this variability has not been explored, in part because available cliff observations are irregular. Here, we systematically mapped and tracked ice cliffs across four debris-covered glaciers in High Mountain Asia for every late ablation season from 2009 to 2019 using high-resolution multi-spectral satellite imagery. We then quantified the processes occurring at the feature scale to train a stochastic birth-death model to represent the cliff population dynamics. Our results show that while the cliff relative area can change by up to 20% from year to year, the natural long-term variability is constrained, thus defining a glacier-specific cliff carrying capacity. In a subsequent step, the inclusion of external drivers related to climate, glacier dynamics and hydrology highlights the influence of these variables on the cliff population dynamics, which is usually not a direct one due to the complexity and interdependence of the processes taking place at the glacier surface. In some extreme cases (here, a glacier surge), these external drivers may lead to a reorganization of the cliffs at the glacier surface and a change in the natural variability. These results have implications for the melt of debris-covered glaciers, in addition to showing the high rate of changes at their surface and highlighting some of the links between cliff population and glacier state.