This is also very important in many global illumination techniques which requires sampling the hemisphere around a point.
As case in point a simple Monte Carlo path tracer uses this to sample random directions, see e.g. [0]. In addition to this there are many 'tricks' one can use to reduce the variance/noise using a differently weighted sampling (to help the converge of MC).
A nice technique to get a cosine-weighted hemisphere sampling is to generate a random point on the unit disc and then project them onto the (hemi)sphere.
Seems somewhat relevant here.
[0] https://en.wikipedia.org/wiki/Path_tracing#Algorithm