Collaborative projects designed to generate research evidence involve knowledge exchange (KE) which hinges on the expectations and practices within the collaborating organisations. Existing literature about academic-police collaborations, and why they break down, has largely focused on different knowledge agendas, research timeframes and organisational cultures. In contrast, we offer a case example that attends to the failures in KE stemming from the basic logistics, legal administrative procedures, and buy-in associated with collaborative projects. These failures in KE meant our planned study involving innovative research methods to assess (de-)escalation in use-of-force incidents using policy Body Worn Video was not feasible, and required adaptation to find another source of data to successfully complete a project. This case offers many lessons learned that can inform future collaborative projects, as well as expanding upon the simplified way KE is often conceived. First, formal agreement from both organisations and the co-production of a project do not imply it is feasible. Second, KE needs to account for the socio-technical networks that are required for successful collaboration and the range of skills this requires to achieve agreed aims. Third, collaborations should recognise the multidimensionality of organisations and expand networks to optimise flexibility and adapt to change so that projects are resilient.