Longwall mining of coal below massive strata now has a 50 year history, and much of it has been problematic. For the first 30 years, there was much experimentation, with some successes, and many failures due to lack of understanding about how such massive strata behaves during caving, and how this differs from the more conventional and successful longwall mining beneath softer strata, which caves readily. This has been compounded over the last 20 years as mines extract wider faces and thicker seams in a single pass, often at quite shallow depths. Today, China is leading the way with thick seam longwalls under massive strata in shallow conditions, though not without problems. Determination of the required capacity for longwall roof supports in such conditions is still not adequately understood, and overloading of supports remains a common problem in modern mines. The authors’ view is that mathematical modelling to determine support requirements can only succeed if the model reflects real observed behaviour in such conditions. There is an extensive body of technical literature documenting observations of what actually happens. This paper draws on past experience in many different countries to categorise common themes, and then proposes in simple forms the basis for real behaviour. Future modelling should be advanced on this basis. Such models need to incorporate a combination of multiple failure modes, including tensile, bending and shear, which may produce large rock blocks. These may move laterally, vertically, or in rotation depending on the loads which act upon them. Large blocks may impose substantial loading on the face, the powered supports or on the gateroad pillars. The problems associated with large block formation may be mitigated by preconditioning to break up the massive strata before problems occur, so that manageable blocks are formed. In the United States of America and Australia, the focus has been on hydraulic fracturing. In Europe, China and South Africa, explosives have more often been used.