Evaluating the temporal stability and risk of unsaturated soil slopes during rainfall is essential for early warning and emergency response to landslides. However, limited research has been conducted on the transition timing of sliding mechanisms, instability/failure time and the integration of sliding consequences into quantitative risk assessment. In order to extend the research in this field, the Random Finite Element Method (RFEM) is used in this paper to investigate the influence of spatial variability of hydraulic properties (related to the fundamental parameter porosity) on the temporal stability and risk of soil slopes subject to rainfall. The findings indicate that the advancing speed of the wetting front is more rapidly in zones with low porosity than that in zones with high porosity. As rainfall progresses, the sliding mechanism of the slope shifts from deep sliding to shallow sliding. The homogeneous case tends to underestimate the rise in groundwater levels, leading to an overestimation of slope stability. Hydraulic boundary conditions significantly affect slope stability, making it crucial to consider horizontal (or near the toe of the slope) drainage conditions in practical applications. Additionally, the time of instability/failure predicted in the homogeneous case may be delayed compared to the actual conditions. Both probability of instability/failure and risk increase with continued rainfall. Compared to scenarios where the spatial variability of internal friction angle is not considered, the probability of instability/failure and risk will be higher when the spatial variability of internal friction angle is additionally considered. Risk-based assessment can define the risk levels, reflecting the severity of sliding consequences. Furthermore, the Malin slope failure record from the Chibo region of India is used to validate the effectiveness of the proposed approach. The probabilities of slope failure align well with actual observations, and the risk-based assessment provides additional information into the Malin landslide. This paper proposes a general model for studying the performance of heterogeneous slopes subject to rainfall, providing valuable guidance for landslide early warning systems and the scope and timing of emergency measures taken.