Eruption forecasting is a major goal in volcanology. The intrinsic complexity and non-linearity of volcanic systems has recently led to an intertwining of deterministic to probabilistic eruption forecasting methods. However, many volcanoes often pass states of non-eruptive and non-magmatic unrest for various periods of time. Logically, but unfortunately, forecasting hazards related to non-magmatic unrest is too often overshadowed by eruption forecasting. Nevertheless, volcanic hazards related to non-magmatic unrest i) can be highly destructive (e.g. phreatic eruptions), ii) can lead into magmatic and eventually eruptive unrest, and iii) can be more difficult to predict, for various reasons: (1) many poorly studied volcanoes are in a state of non-magmatic unrest, passing unnoticed eventually leading to hazardous surprise events (e.g. El Chichón-1982, Pinatubo-1991, Chaitén-2008), (2) the duration of a state of non-magmatic unrest can be highly variable, (3) the cause and type of hazardous event and the locus of occurrence can be highly variable, even for a single volcano, (4) non-magmatic hazards can be caused by factors external to the volcano itself, and (5) monitoring approaches should not be limited to the usual seismicity-ground deformation-gas, and therefore requires additional monitoring setups. In this study we summarize how to recognize non-magmatic unrest and precursors for related hazards. We propose an event-tree for non-magmatic unrest, which aims to cover all major hazardous outcomes, and which could become the structure for future probabilistic non-magmatic hazard assessments.