Why we get dark diffused sphere in places where normals a almost perpendicular to the plane ?

[tempdeltavalue]

As I understand these ray just should be bounced "to the viewer" without recursive colour decreasing ..

Thank you !

[tempdeltavalue]

I get such sphere (in glsl), I assume what something wrong with random, "upper hemisphere" random vectors point in direction opposite to surface , or addition to normal cancel out scattered ray

https://github.com/tempdeltavalue/visual_temp/blob/main/visual_temp/glsl/ray_tracing/fragment.glsl