I find both of your "concerns" problematic. First on the M/M study, statistical significance as you well know does not rely upon a large numerical difference between drug and controls. It is a measure of probability, the probability of a non-random effect. Thus, all they need to show is that there are at least some controls with at least a symptom compared to experimental group within a large enough N. This brings us too your second objection of there not being a large enough N for a binomial measure to carry much information unless the N is very large. Again the N need not be large if the effect is substantial, which we have ever reason to believe it will be based on the data we already have. Also, take a look at the Remdesivir trials. If that is the statistical baseline standard for effect, I share very little of you concerns. IMHO, of course.