Miller and Brody argue that the notion of clinical equipoise is fundamentally misguided. The ethics of therapy and the ethics of research are two distinct enterprises that are governed by different norms. They state, “The doctrine of clinical equipoise is intended to act as a bridge between therapy and research, allegedly making it possible to conduct RCTs without sacrificing the therapeutic obligation of physicians to provide treatment according to a scientifically validated standard of care. This constitutes therapeutic misconception concerning the ethics of clinical trials, analogous to the tendency of patient volunteers to confuse treatment in the context of RCTs with routine medical care.”  Equipoise, they argue, only makes sense as a normative assumption for clinical trials if one assumes that researchers have therapeutic obligations to their research participants. Further criticisms of clinical equipoise have been leveled by Robert Veatch  and by Peter Ubel and Robert Silbergleit. 
The conversation here is very interesting, but it is impt to get the definition of equipoise right. The operational definition is not whether the researchers themselves are uncertain, but whether there is meaningful uncertainty, or observed variation, among the community of practitioners, which in this case might be the policy makers, and possibly researchers. Freedman's contribution was to eliminate the concept of individual researcher (or team) uncertainty from the mix, as long as there is meaningful disagreement in the community. Now, the second question is what is teh meaningful disagreement about? If it is about allocations of money to qualitatively different health interventions, then that should be the randomization, if indeed it is possible. If there is little doubt about the efficacy of a given allocation, or intervention, it probably shouldn't be randomized against not giving that intervention, although that depends on background conditions. Randomization to a suboptimal state can be justified depending on the counterfactual in that area. So this is indeed a complicated question, and parallels w/medicine aren't perfect. There is something to be learned from the thinking that has gone on in medicine, but it has to be correctly framed. But the better medical parallel to development is the area of systems or quality improvement, which even in medicine can be very context-dependent.