This function does indeed calculate the average value of all the z-coordinates, and will prevent the error occuring when there is no duplicate z-coords.

However, this does produce a different result that the current ox implementation.

I created a simple test here.

When you provide a list of points without a duplicate z-coord, the current ox implementation breaks. But when a duplicate z-coord is present, it doesn't break, but also doesn't calculate the average. It's either the median value, or some kind of biased average.

@thelindat @DokaDoka The implementation in this PR would produce the average z-coord and prevent the code breaking, but I am unsure if average is the desired result.

Edit: On further review, it looks like current implementation is just picking the most common z-value. See #723 for alternative fix, which maintains current implementation.

Related to this Discord discussion.