The depth and screen textures both become texture arrays when the main render target is multiview. Consequences for the shader code, incl. user-provided? (one cannot probably just write the GLSL code 'texture(SCREEN_TEXTURE, uv)' anymore, once we fix up the shader generator so the sampler is correctly a sampler2DArray...)

Source compatibility and portability concerns: e.g. if SCREEN_TEXTURE is simply not a sampler2D anymore in multiview mode, what if users want to write custom materials and effects that work both in regular and in XR mode? Also is this sensible even; how would the shader know that now it has to do things twice, first by sampling layer 0 and then repeating the same for layer 1? More uniforms?