Materials under an instancer's Prototypes are dropped in the Scene Index pipeline

[robp-sidefx]

The UsdImagingPiPrototypePropagatingSceneIndex seems to type-erase all the Hydra prims under an instancer except for the meshes that are actually instanced. This includes materials (i.e., the renderer never gets access to them), which wasn't the case previously.

The following USD renders a blue sphere in "Hydra 1" and a grey sphere in "Hydra 2"

#usda 1.0
(
    endTimeCode = 1
    framesPerSecond = 24
    metersPerUnit = 1
    startTimeCode = 1
    timeCodesPerSecond = 24
    upAxis = "Y"
)

def PointInstancer "instancer"
{
    point3f[] positions = [(0, 0, 0)]
    int[] protoIndices = [0]
    rel prototypes = [</instancer/Prototypes/sphere_USED>]

    def Scope "Prototypes"
    {
        def Xform "sphere_USED" (
            prepend apiSchemas = ["MaterialBindingAPI"]
        )
        {
            rel material:binding = </instancer/Prototypes/materials/blue>
            def Sphere "sphere"
            {
                double radius = 1
            }
        }

        def Xform "sphere_UNUSED" (
            prepend apiSchemas = ["MaterialBindingAPI"]
        )
        {
            rel material:binding = </instancer/Prototypes/materials/blue>
            def Sphere "sphere"
            {
                double radius = 1
            }
        }

        def Scope "materials"
        {
            def Material "blue" 
            {
                token outputs:mtlx:surface.connect = </instancer/Prototypes/materials/blue/mtlxstandard_surface.outputs:out>
                def Shader "mtlxstandard_surface"
                {
                    uniform token info:id = "ND_standard_surface_surfaceshader"
                    color3f inputs:base_color = (0, 0, 1) 
                    token outputs:out
                }
            }
        }
    }
}

I definitely don't mean to go so far as to say "the old way was right, the new way is wrong", but I do flag there being a change (which, unfortunately, got noticed by one of our users).

[jesschimein]

Filed as internal issue #USD-10562

(This is an automated message. See here for more information.)