With solar cells reaching 26.1% certified efficiency, hybrid perovskites are now the most efficient thin film photovoltaic material. Though substantial effort has focussed on synthesis approaches and device architectures to further improve perovskite-based solar cells, more work is needed to correlate physical properties of the underlying film structure with device performance. Here, using cathodoluminescence microscopy coupled with unsupervised machine learning, we quantify how nanoscale heterogeneity globally builds up within a large morphological grain of hybrid perovskite when exposed to extrinsic stimuli such as charge accumulation from electron beams or milder environmental factors like humidity. The converged electron-beam excitation allows us to map PbI2 and the emergence of other intermediate phases with high spatial and energy resolution. In contrast with recent reports of hybrid perovskite cathodoluminescence, we observe no significant change in the PbI2 signatures, even after high-energy electron beam excitation. In fact, we can exploit the stable PbI2 signatures to quantitatively map how hybrid perovskites degrade. Moreover, we show how our methodology allows disentangling of the photophysics associated with photon recycling and band-edge emission with sub-micron resolution using a fundamental understanding of electron interactions in hybrid perovskites.