This is the old school joint-based facial rig for game I want to present. The point was making the rig only once, the pose library only once and skin weights only once and transfer these for a new characters, as I don't want to have to replicate my work over and over.
So basically cage solve all these problems (apart from skin weights obviously, that we reach thanks to universal topology). The reason that is just because of how rig is layered. When the cage morphs, the locators get driven and when the locators move they move the offset rig as well and then the offset rig drives final game joints. The offset rig helps animators to add something more to animation.
The nice thing is that I use only simply maya tools, nothing fancy scripts or expressions.
I was very inspired by lecture of Jeremy Ernst https://www.gdcvault.com/play/1014366/Fast-and-Efficient-Facial-Rigging.
If you want to see this rig in action you can check my next artwork 'Facial Animation Sharing': https://www.artstation.com/artwork/oANoEB
You want to make rig easy to animate - it's self-explanatory, but also hard to do. So as you can see we have one slider to get the pose that i want and then on top of that the offset rig, which allows the animator more control, but in a unique way.