Background From January 2015 to December 2016, the health authorities in Burundi piloted the inclusion of child nutrition services into the pre-existing performance-based financing free health care policy (PBF-FHC). An impact evaluation, focused on health centres, found positive effects both in terms of volume of services and quality of care. To some extent, this result is puzzling given the harshness of the contextual constraints related to the fragile setting. Methods With a multi-methods approach, we explored how contextual and implementation constraints interacted with the pre-identified tracks of effect transmission embodied in the intervention. For our analysis, we used a hypothetical Theory of Change (ToC) that mapped a set of seven tracks through which the intervention might develop positive effects for children suffering from malnutrition. We built our analysis on (1) findings from the facility surveys and (2) extra qualitative data (logbooks, interviews and operational document reviews). Findings Our results suggest that six constraints have weighted upon the intervention: (1) initial low skills of health workers; (2) unavailability of resources (including nutritional dietary inputs and equipment); (3) payment delays; (4) suboptimal information; (5) restrictions on autonomy; and (6) low intensity of supervision. Together, they have affected the intensity of the intervention, especially during its first year. From our analysis of the ToC, we noted that the positive effects largely occurred as a result of the incentive and information tracks. Qualitative data suggests that health centres have circumvented the many constraints by relying on