I suggested something smilliar in the forum a few weeks ago. A new parameter for face recovery strength users can set. Then a second slider for face detection sensitivity
This allows faces to be reconstructed as strong/good we know as it is, but by adjusting the sensitivity downwards, only easily recognizable faces would be reconstructed. In this way, no attempt would be made to reconstruct bad faces that are difficult to recognize in images, which is often the case with wide shots and leads do disfigures faces.
