Motion patterns in toolbars are strict: you have to start at the top and drill down. This seems like a hindrance, but it means that the motions develop into stronger muscle memory. If you know the names of what you are looking for, you can usually develop the entire shortcut pattern through everyday use, without setting aside practice time(as in a setup like vim or emacs, where you aren't given sufficient prompting to discover and train new interactions automatically).

Ribbons surface more elements to browse in a freeform context, which is correct if you need to discover features...but also conflicts with the goal of a toolbar to be a thin layer over the shortcuts.