We explore some recurring socio-technical problems encountered in the development of infrastructure for sharing and re-using data across sites and social scales for eHealth research. We link these problems to contradictions between underlying assumptions about data as a commodity whose reuse is not compromised when it is extracted from the context in which it has been captured, and the reality of data as entangled with, and constituted through, local practice. To illustrate these problems, we draw on the experiences of a number of HealthGrid projects developing infrastructures for data sharing and reuse, and trace the strategies that have evolved to address them. These experiences problematize the "one size fits all" model initially adopted by HealthGrids, and highlight the need for design and development strategies that are able to engage with local needs and thereby ensure that the technical infrastructure is properly aligned with the human infrastructure it is supposed to support.