If the best available means to perform an experiment carries some risk, it could still be entirely rational to do it rather than forfeit the knowledge gained from the experiment.
For example take the famous mask debate. It could easily be solved by having volunteers willing to stand in a room with people with covid at various distance, each using randomized masks/no mask. There would be plenty of volunteers for such a study but there's no way it would be approved.
The FDA doesn't count lives lost due to inaction and slow approval of new drugs and treatments. As Munger always said "show me the incentive and I'll show you the outcome." By any rational calculus, that one Thalidomide win by the FDA has caused incalculable death, pain and suffering by pushing out the timeline on not only recently discovered cures but all those built on top.
Imagine for example the number of lives saved if GLP-1 was purchasable over the counter in the 1990s when it was first discovered.
If the best available means to perform an experiment carries some risk, it could still be entirely rational to do it rather than forfeit the knowledge gained from the experiment.