With the quantized mesh representation for terrain, I wonder if the system supports drawing concavities on the globe's surface.
Potential issues from a quick glance:
- implementation of TerrainData's interpolateHeight method
- implementation of TerrainData's upsample method
- occlusion based optimizations within Cesium (they may not be suboptimal now, or require gross inputs to "make things look right")
What do you all think?