On the scenery side at least their main problem is the lack of LODs. Most have just a LOD0, which is their highest detail. Others also include either a terrain texture, a elevation mesh or both, and that alone can be several GB (can be even heavier that the complete 3D model of the airport buildings for instance). In particular the ground texture appears to load in one go, while the Bing maps textures load in smaller tiles.
In any of those cases several gigas are involved in the data transfer (to load or unload assets in memory) so that takes a bit more time than usual.
Cheers