Are gl_BaseInstance/Vertex worth it for WEBGL_draw_instanced_base_vertex_base_instance?

We need to polyfill these all the way until GLSL460 and/or ARB_shader_draw_parameters.[[1]]
That polyfill might at-worst incur a uniform upload per draw call. I'm worried about this overhead on all our pre-4.6 systems. Interleaving draw calls with uploads seems like a bad idea for many-draw-call workflows.

[1]: https://www.khronos.org/opengl/wiki/Vertex_Shader/Defined_Inputs

Start with the issue discussion and the linked Khronos Vertex Shader/Defined Inputs reference. Determine whether polyfilling gl_BaseInstance and gl_BaseVertex through GLSL 460 or ARB_shader_draw_parameters is feasible without unacceptable overhead for pre-4.6 systems. Done means reaching and documenting a decision on the polyfill approach and its draw-call performance tradeoffs.

Written by the indexing model from the issue text.