Using opacityscale with surface results in ordering by addition order, not depth

Hi all, been using plotly with R for a while now and really loving it - thanks for assembling such a phenomenal program! I'd like to report a bug found while plotting multiple surface objects, which causes them to layer in the order they're added rather than the "depth" at which they should really render in the figure. See this codepen for code and examples, which are also rendered below as screenshots.

I previously posted in ropensci/plotly, but have since learned enough Javascript to assemble the codepen and the issue persists, making me think it's not an R problem. There's also no discussion of the new "opacityscale" feature there and a lively discussion here, making me hopeful that more people will have some idea how I might fix this if it isn't a bug.

This is great:

This is less great (same figure rotated to look from below):

It seems that the surface that renders always on "top" is the one that's added last in the data array. If we alter the above code a little bit to render the top layer surface before the middle one, we get some more interesting results:

Plotly.newPlot('graphDivReorderOpac', [data_z1, data_z3, data_z2]); rather than Plotly.newPlot('graphDivReorderOpac', [data_z1, data_z2, data_z3]);

Apologies for the repetitive code in the codepen - I really did learn just enough JS to make this one and will be hiding behind R wrappers in the future.

Start with the linked CodePen and the opacityscale surface examples, comparing the two trace orders and the resulting rendering from above and below. Trace the surface rendering path responsible for depth ordering; done means overlapping surfaces render by depth rather than data-array order, while preserving the working non-opacityscale example.