Is this only breaking the sample scenes? If you make a new scene with a brand-new super text mesh object, does it work fine? Or if you add a new Super Text Mesh to a scene in your game?
And the typing input scene doesn't have a caret, it's mostly just to demonstrate how to use tags.