It seems in the real world that one of the important aspects of reproducibility is to remove, or at least in some fashion account for, bias. First and foremost, this may be investigator bias. However, there are other biases of importance. Examples in clinical research abound, including those arising from study site, method of reimbursement, and competing interests.
Given the relevance, and potentially large effect, of these biases, which might be reduced through efforts to reproduce findings, it seems that a conclusion that down plays the value of reproducibility may be premature.