I'm not saying this is always done, but rather it shows such theories are falsifiable in the sense of Popper.
Let's take an example I have been contemplating lately: pruny fingers. When my baby goes in the water, his fingerpads and toes turn wrinkly. There is a supposition, part of the "Aquatic Ape Theory" that is a famous example of an evolutionary just-so story, that this is an adaptation to allow better grips in the water. Sounds good to me - but how could we test this claim?
We could start with physical tests (does grip actually improve when your fingers are pruney?), but we immediately run into trouble. A grip on what? In what sort of conditions? Did these conditions exist in the adaptive context that produced this phenotype? We have no idea.
There might be rare circumstances where we can observe the coincidence of environmental and phenotypic isolation (e.g. cattle herders who are lactase persistent), but the vast majority of features will be lost against the general evolutionary background of our species (e.g. everything separating humand from chimps, both genetically and environmentally, is one confused ball). But even those rare cases are still inferential; the inference seems strong and parsimonious, but its still, at the end, "just so". We can never prove causality.