They were blinded to patient data but not to the results on any efficacy IAs that were performed. And we know at least one was. And very likely more.
The assertions here that being blinding implies the change is fine shows a basic lack of understanding. What is key is that they change cannot be made based on influence from data within the trial. Changing the endpoint after knowing the original endpoint failed is more than a bit questionable.
The RAs agreed the trial changes were safe and ethical for the patents. Many trials are intentionally not designed to establish proof worthy of approval, so it is not the RAs job to stop trials for that reason.
OS in a randomized controlled study is the gold standard. The FDA has many guidance notes where they note that OS vs non randomized controls has issues.