I agree that giving HCQ to a very sick patient is a very bad idea, because his heart is already going to be in bad shape. But the protocol that would be useful, would be to give it to him long before he had any such bad effects. If it does work, he would never get to that stage. The stage in which it might work, would be from first symptom to hospitalization. This is what some doctors are calling stage 2: When they would normally send patients home and tell them to come back if they have shortness of breath.

I've looked at an analysis of the VA study, and it appears to be completely useless: The HCQ arm of the study was only the sickest patients, while the no-HCQ arm was patients who did not get that sick. Of course it looked like HCQ did harm: When you just give an antiviral drug to the sickest patients it has no effect, but they were the high-fatality patients, drug or no drug! I doubt you can get valid patients-to-treat or patient-to-harm numbers from that mess.