* The copyleft license ODbL probably will mean that there never will be a huge adoption. The situation simply is different with a geo database compared to software or one art piece.
* Vandalism very easily goes unnoticed, as there is no review at all for edits. If something that went wrong is found, reverting is very hard to do, often left to some expert group that only does that.
* Tags indicating when things were last checked are hardly in use. So the database contains tons of information that might not be true anymore or might never have been true.
* The data model lacks persistency, the tags of a shop might go from a node to the building area it's in or otherwise effectively loosing the history. It's very hard to find out where something really came from (editor and source).
* The data model also lacks a good way of dealing with areas (especially ones with holes in them), multipolygon relations are a hack barely supported in editors that often lead beginners to feel overwhelmed. On the other hand there are no curves, just straight lines, because bezier lines are supposed to be too confusing for beginners.
* All the data is in one huge database, not split up into layers.
* There are hardly any volunteers in America. It's an import desert.