I do not need cube texture now. If it is hard to implement, do not do it. Shadows from point lights will be very expensive especially on mobile GPUs. Now for me have a higher priority game elements than effects.
Shadows. It depends on what you mean static scene. I think, it is very big difference. Shadow map from light can be updated only, when some object in view frustum of the shadow projector or projector performs a transformation. In other cases can be the same and camera can move in the scene without updating shadow map. I think it would be enough calling onTransform() method when is called some of the transformation methods like translate(...), rotateX(...)... and input is not zero. Other optimization can be render in small part of texture only. If object, that performed transformation is only one, rendering can be perform in the part of texture where AABB of object occupies.