How Layer 3D To Feature Class (3D Analyst) works
The Layer 3D To Feature Class tool will apply certain 3D properties defined by the layer to the input features and write them out to a new feature class. The input feature type coupled with the 3D properties must allow representation as a multipatch or extruded points.
If a feature can't be converted into a multipatch, it will not be written to the new feature class. For example, it's possible that some points are symbolized by 3D markers and some are not. Only those symbolized by 3D markers can be converted.
If you have polygons with heights (z-values) and want to view them in ArcGlobe without being rasterized and draped on the globe surface, they need to be converted to multipatches. The Layer 3D To Feature Class tool can be used to perform this conversion.
Note:
Often, it's best to have on-the-fly properties applied by a layer remain that way. You wouldn't want to treat the symbolic representation of a feature literally as a feature. On the other hand, this can be useful for bootstrapping a 3D database. Take, for example, a collection of extruded building footprints. These are derived from polygons attributed with height. If you save the extruded representation as multipatch geometry, you have the opportunity to refine the geometry through use of third-party editors. Google SketchUp software is one such editor that can be used to read, edit, and write multipatches. With this tool, you can add pitched roofs and textures.