Operating system: NixOS 19.03, Linux 4.19.36
Graphics card: NVIDIA 970 GTX
Broken: 2.79, 2.80 git (93cabce3a156a07a76c26d7f3e78c3bc3212f623)
Short description of error
Using the Principled BSDF's "Random Walk" subsurface scattering method, if the "Subsurface" input is set to a low, non-zero value (e.g. < 0.01), you get a very dark looking object. For example, the following image shows three Suzannes with the same material except a different "Subsurface" value, from left to right 0, 0.0001, and 0.1:
The Blender manual (https://docs.blender.org/manual/en/dev/render/cycles/nodes/types/shaders/principled.html) somewhat cryptically states that the Subsurface input acts both as a mix factor between Diffuse and SSS *and* as a multiplier for the subsurface radius, but either way this doesn't seem like expected behaviour. It doesn't occur with the Christensen-Burley method.
Exact steps for others to reproduce the error