How can I prevent Scatter3 and plot3 points from occasionally getting rendered in front of a surface plot despite being physically behind it?

I'm trying to make an animation of the publicly available cataloged objects in Earth orbit, but occasionally, and seemingly randomly, they are not rendered properly. Here are two cropped screenshots from the simulation that show what I mean:

And the very next frame:

In the second image, ALL points (plotted with plot3() in this instance) are rendered in front of the spherical surface plot representing the Earth, even though half of them are behind it. Maybe it's more clear to see a short sample of the actual animation: https://gfycat.com/tinygreenbeauceron The effect occurs halfway through, as the camera movement changes. However, I don't think the issue is exclusively due to the camera movement - the rendering issue is actually much worse if I use scatter3() instead of plot3(), and occurs in most of the frames of the animation. I would prefer to use scatter3() as it allows me to change the alpha values, whereas plot3() does not.

I'm not sure if I can paste code here to reproduce a "minimum working example" because I can't figure out the pattern that causes this to happen and I'm not sure how to reproduce the effect with a simpler setup. I did try, but was unable to observe the same effect. The code that does the actual plotting is pasted below, but of course it can't be run without the data from the rest of the simulation, which is too large to provide:

f1 = figure(1);
set(f1,'units','normalized','outerposition',[0 0 1 1]);

if ~strcmp(vfname,'null')
    v1 = VideoWriter('2019-06-14_catalog_animation','MPEG-4');
    v1.Quality = 100;
    v1.FrameRate = 60;
    open(v1);
    nframes = numel(CamVA);
else
    tvec_days = tvec_days(1);
    nframes = 1;
end

for ii=1:nframes

    % All objects:
    pl(1) = plot3(r_all(1,:,ii),r_all(2,:,ii),r_all(3,:,ii),'w.','Markersize',0.1);
    hold on
    plot_Earth_km(eq_pts);
    hold off
    campos(rCam(:,ii))
    camtarget([0;0;0])
    camva(CamVA(ii))
    camproj('perspective')
    
    grid off
    set(gca,'xtick',[])
    set(gca,'xticklabel',[])
    set(gca,'ytick',[])
    set(gca,'yticklabel',[])
    set(gca,'ztick',[])
    set(gca,'zticklabel',[])
    set(gca,'XColor','k','YColor','k','ZColor','k')
    set(gca,'Color','k')
    
    % Set the axis limits to ensure the camera is inside the axes:
    xLimits = max(abs([get(gca,'XLim'),rCam(1,ii)+R_E]));
    yLimits = max(abs([get(gca,'YLim'),rCam(2,ii)+R_E]));
    zLimits = max(abs([get(gca,'ZLim'),rCam(3,ii)+R_E]));
    axis([-xLimits,xLimits,-yLimits,yLimits,-zLimits,zLimits])
    set(gcf,'renderer','openGL')
    drawnow
    
    if ~strcmp(vfname,'null')
        frame = getframe(gca);
        writeVideo(v1,frame);
    end
end

if ~strcmp(vfname,'null')
    close(v1);
end

The plot_Earth_km() function generates the surface plot of the Earth, after computing the correct lighting and generating a texturemap to apply to the surface plot. It's a very large function but the part that does the actual plotting is nothing special:

function [ax,Gh] = plot_Earth_km(varargin)
% Plots the Earth with a textured map to scale in kilometers. 

% Parse inputs - only 'eq_pts' (number of equatorial points) is required
[...]

% Define the surface of the sphere
RE_eq = 6378.137; %[km] Earth equatorial radius (WGS 84)
RE_pol = 6356.752314245; %[km] Earth polar radius (WGS 84)
[Earthx,Earthy,Earthz] = ellipsoid(0,0,0,RE_eq,RE_eq,RE_pol,eq_pts);

% Generate the texturemap 'cdata' with accurate lighting
[...]

Gh = surf(ax, Earthx,Earthy,-Earthz, 'FaceColor','texturemap',...
        'CData',cdata, 'EdgeColor','none');

If I stop the process on one of the frames where the issue arises, and manually change the camera settings by using the rotate3d tool or zoom tool, the points get correctly plotted behind the Earth (but for some reason the camera zooms WAY out and I have to zoom back in). I have been trying to programmatically do the same thing before capturing the frame, using zoom() and rotate3d() to change the view and then change it back, but no automated actions seem to "refresh" the render the same way that using the buttons in the toolbar of the plot3 GUI does. Manually using the "pan" tool does not refresh the render.

Thanks very much for any suggestions or help!

[Walter Roberson]

A few weeks ago someone else was mentioning depth problems somewhat similar. Unfortunately at the moment I do not recall enough keywords to pull out the Question.

[Nathan]

Thanks for the heads up. I'm trying to search for the post you're referring to but I haven't had any luck either.

[Walter Roberson]

If you happen to have the right toolbox it could be interesting to try the volume viewer test mentioned in

https://www.mathworks.com/matlabcentral/answers/471279-3d-structure-not-displaying-correctly-volumeviewer-example-3d-data-does-not-display-properly

As that would give us information about whether this is a general rendering problem with your system or an issue only for scatter3

Speaking of which, use

opengl info

to get matlab to display information about which graphics card and driver you are using. I have noticed that it is not uncommon for people with AMD graphics cards to be 2 or more years out of date on their drivers.

[Nathan]

I don't think I have the right toolbox. My version of Matlab does not recognize the command "volumeViewer" and it cannot find the directory requested by those commands:

>> load(fullfile(toolboxdir('images'),'imdata','BrainMRILabeled','labels','label_001.mat'));
Error using load
Unable to read file 'C:\Program Files\MATLAB\R2015b\toolbox\images\imdata\BrainMRILabeled\labels\label_001.mat'.
No such file or directory.

I realize it might be relevant to note that my version of Matlab is 2015b. I can't seem to add a release version in my question - the option is grayed out.

My computer has an Nvidia graphics card and the drivers were from May 2019, but I updated them just in case and ran the script again but saw no change.

[Walter Roberson]

(I have reported the glitch about entering version information.)

[Walter Roberson]

I do not appear to find any relevant (public) Bug reports

[Nathan]

Thanks Walter. I've been trying lots of different things to try to get it to render properly but now I'm running into another issue - large "rips" for lack of a better word. I rescale the axes so that the camera position is significantly inside them, but for whatever reason it seems that large chunks of the screen occasionally render outside the axes anyway, creating white triangles: https://gfycat.com/perfumedunrealisticafricangoldencat

I also added a large magenta dot at the center of the Earth just to make it more obvious when rendering issues occur, since it should never be visible.

[Walter Roberson]

Odd. I do not have an explanation.

Can you attach your code and data for testing?

[Nathan]

The limited file sizes on attachments makes it difficult to give you exactly what I have but I did my best to simplify everything while keeping it "effectively" identical. You'll have to propagate the objects from their initial positions on the first run through the script, but if you save them to the workspace it should skip it on subsequent runs. Here is a video of the resulting animation using exactly the code attached to this comment: https://gfycat.com/chubbygorgeousaustralianfurseal

No flashing triangles this time (I have no idea how to replicate that again - it doesn't even do it with my "non-simplified" code anymore, but I'm not sure what I changed). The magenta dot and objects in Low Earth Orbit are still visible when they shouldn't be, but it appears that objects in the geostationary ring are correctly hidden at the beginning. What's more, I discovered that increasing the markersize of the magenta dot from 10 to 15 will cause it to be hidden.

Thank you for offering to look into this, I really appreciate your help.