We have relied too heavily on the notion of only analyzing nodes marked as "samples". While this is likely the most common thing to do, we should not be requiring this! (Unless we want to be less general than we could be). Our naive tests are only working for nodes marked as sample -- we should uncouple the testing methods from the sample-ness of nodes before moving on with, e.g., #100.
We have relied too heavily on the notion of only analyzing nodes marked as "samples". While this is likely the most common thing to do, we should not be requiring this! (Unless we want to be less general than we could be). Our naive tests are only working for nodes marked as sample -- we should uncouple the testing methods from the sample-ness of nodes before moving on with, e.g., #100.