I think the .unit class could be done away with if he used .ingrid > * in its place. Of course, that might mean rewriting some of the CSS to account for the change in specificity in that rule.
I believe this would have performance implications (css selectors are read "backwards" by browsers, so this would mean every element on the page is examined to see if it's inside a .ingrid class). Not always, but definitely could for a large/complex page, and since this is a framework that will be used in all sorts of situations it's best that the author stays on the conservative side.