Well‐controlled landscape experiments have played key roles in advancing fragmentation science, but such experiments are costly and may not be possible in many ecosystems – including the long‐inhabited landscapes typical of many developing countries. In such contexts observational studies of pre‐existing forest patches may offer valuable insights, but these bring other challenges including the non‐random location of patches, the heterogeneous matrix between patches, and patch‐specific management practices that may influence forest community composition. This paper argues that sacred natural sites might provide a middle ground between experimental and observational studies, allowing for more rigorous mensurative fragmentation experiments in long‐inhabited multi‐functional landscapes. To illustrate this potential, we analyze the drivers of productivity in ‘church forests’ across northern Ethiopia. These forest patches conserved by followers of the Ethiopian Orthodox church provide ample variation in area, edge, and surrounding matrix characteristics. Church forests also provide variation in long‐term forest community composition, elevation and rainfall. Finally, unlike most observational studies, church forests offer a relatively stable institutional structure, including longstanding religious norms, allowing researchers to control for some heterogeneity in human influences. By combining remotely sensed data on church forest patches (n = 2558) with field data on church forest tree species composition (n = 27) and social survey data on church forest management practices (n = 145 respondents in 6 church communities) we show how ecological and anthropogenic factors influence church forest productivity. Like experimental patches, church forest productivity increases with size and decreases with amount of edge; productivity also increases with rainfall and increased tree species diversity within a given patch. But there is also evidence that church forest productivity and species composition are both affected by human management rooted in longstanding religious norms. Findings highlight how studies in sacred natural sites systems might help understand relationships between forest productivity, species diversity, and human management in long‐fragmented landscapes.