The disk BSSRDF kernels have a correction method to "adjust the pdf" when branched-path isn't being used, because in path tracing not all shader closures are evaluated on each sample - just one is selected.

The random walk kernel doesn't have such a correction, which results in dark renders when multiple BSSRDFs are included (mixed) in the shader.

This code change makes the "dark renders" problem .. go away. But I really don't know if this is the right way to achieve that - and if it is, whether this is the right way to code it. I have attempted to follow the pattern of the existing code and I've had a little go at understanding https://sites.fas.harvard.edu/~cs278/papers/veach.pdf. I think all my code does is reverse the weight applied to throughput in shader_bssrdf_pick in kernel_shader.h.

The attached blend

uses a checker texture to compare a single sss shader vs a mix shader with two sss shaders duplicated from the first.

The renders show that the problem goes away when branched path is used. It also goes away if both shaders are changed to one of the disk sss methods.