There's a difference between focus groups and user research. Microsoft primarily uses user research.
Teams always design their own features using their own intuition, their own goals and inspiration, adherence to the Design team's guidelines, and analysis of the competition. They'll then seek out real people to test designs - recruiting thousands of real people of all skill levels, areas of expertise, segregated by platforms they currently use, etc. They'll be presented with prototypes or real implementations, they'll be asked to do specific things, and members of the teams will observe usually without asking any questions.
The teams then look at all the videos, feedback, etc. and make their own decisions about what it means and determine what works and what needs to be improved.
An unclear arrow changing to say "all apps" is an easy thing that simply makes sense to everyone instantly, and no, it doesn't marginalize the design language. You have to have some pretty arbitrarily strict definition of what a "design language" is if you think that this change is somehow ruining Metro.