Currently: If any of the blendshapes declared in the viseme controller is used as part of an animation, vrc seems to ignore them alltogether, and simply doesn't move the mouth.

This is regardless of the animation playing at all, letalone if the layer has zero-weight or not.

An "obvious" usecase would be changing the behavior from tracked to animation-driven in the Animator Tracking Control.

Ofcourse a second set of blendshapes could be made to support that, but that just seems unneccessary for bloat and performance