This paper presents issues which arose in the conduct of qualitative evaluation research within a cluster-randomized, community-level, preventive intervention trial. The research involved the collection of narratives of practice regarding the intervention by community development officers working in eight communities over a two-year period. The community development officers were largely responsible for implementing the intervention. We discuss the challenges associated with the collection of data as the intervention unfolded, in particular, the disputes over cues to revise and adjust the intervention (i.e. to use the early data formatively). We explore the ethical uncertainties that arise when multiple parties have different views on the legitimacy of types of knowledge and the appropriate role of research and theory in various trial stages. These issues are discussed drawing on the fields of ethnography, community psychology, epidemiology, qualitative methodology and notions of research reflexivity. We conclude by arguing that, in addition to the usual practice of having an outcome data-monitoring committee, community intervention trials also require a process data-monitoring committee as a forum for debate and decision-making. Without such a forum, the relevance, ethics and position of qualitative evaluation research within randomized controlled trials are destined to be a point of contention rather than a source of insight.