> With this approach, how would you measure the effectiveness of the sunscreen when it's been absorbed by the skin

You can eliminate the "can't possibly work" cases much faster and cheaper.

More importantly, it is cheap enough to be always used as a baseline verification when human testing is so expensive that it can only be used as a random sample double-check.

It's like unit testing vs full user acceptance testing. You can and should do both, but the latter isn't for every PR.