More clients, less resources: toward a new conceptual framework for agricultural research in marginal areas