Organic solar cells are thought to suffer from poor thermal stability of the active layer nanostructure, a common belief that is based on the extensive work that has been carried out on fullerene-based systems. We show that a widely studied non-fullerene acceptor, the indacenodithienothiophene-based acceptor ITIC, crystallizes in a profoundly different way as compared to fullerenes. Although fullerenes are frozen below the glass-transition temperature T-g of the photovoltaic blend, ITIC can undergo a glass-crystal transition considerably below its high T-g of similar to 180 degrees C. Nanoscopic crystallites of a low-temperature polymorph are able to form through a diffusion-limited crystallization process. The resulting fine-grained nanostructure does not evolve further with time and hence is characterized by a high degree of thermal stability. Instead, above T-g, the low temperature polymorph melts, and micrometer-sized crystals of a high-temperature polymorph develop, enabled by more rapid diffusion and hence long-range mass transport. This leads to the same detrimental decrease in photovoltaic performance that is known to occur also in the case of fullerene-based blends. Besides explaining the superior thermal stability of non-fullerene blends at relatively high temperatures, our work introduces a new rationale for the design of bulk heterojunctions that is not based on the selection of high-T-g materials per se but diffusion-limited crystallization. The planar structure of ITIC and potentially other non-fullerene acceptors readily facilitates the desired glass-crystal transition, which constitutes a significant advantage over fullerenes, and may pave the way for truly stable organic solar cells.