Problem:
The index metadata loading logic is not well optimized. In short, each time a call to IndexMetadataFactory.queryIndex is made the blocks are fully scanned. For small VDBs there isn't much of an issue because of the block caching logic, but during a vdb deploy we could be constantly re-reading data from the files.

Possible solution:
The IndexMetadataFactory loading strategy should be made aware of the index file per schema strategy so that scans can be made per schema rather than against all index files.