With respect to medication, I find your list not much more convincing than personal testimonials which are themselves not at all informative. Trying to figure out the body by doing a bunch of double blind tests is like trying to figure out a car by listening to how different actions change its sound. You may get some way but you won't have anything much better than a list rules for calculating epicycles.
It is true that the prevalence of 12 person tests which barely correct for confounders, misunderstood p-values and outright making up results to get more positive outcomes is a very bad state affairs for medicine. Explaining in part why results are constantly contradicting each other. But even if it weren't so - without understanding a particular drug's sensitivity to genetics and the particular individuals respective haplotype, large numbers and good methodology will simply not not be good enough.
Not until sensing technology becomes more prevalent will it be possible to say anything about the effects of drugs with confidence. We already for example know that time of day, food content and acidity of stomach, how the drug is taken, consistency, metabolism and genetics can affect toxicity and or efficacy. You cannot have any kind of conversation on drugs without knowing what pathways it interacts with, how it saturates, how and if any haplotypes respond differently, where and how it is metabolized, state of vitals when taken and so on.
But there is good news. The effects of genetics on drugs is one of the easiest success stories one can point to for genomics. Sensors are getting better so that one day we can not just share testimonials but sequences, recorded metabolites, change in vitals, oxygenation, hormones etc. to finally be able to form proper correlations. It is true that tests can very much also include this type of data but at that point I don't think the lab experiments would even be necessary. A person could simply randomly place sugar capsules, do the battery of tests and then measure for absorption themselves.
It is for this reason I believe that personalized crowdsourced testimonials backed with hard real data will someday outpace the current lab based/phase 4 method of testing effectiveness.