In the process, we might lose some good "short_description", "website" and "source_code" attribute data for multi-grant projects, but I don't see a way to reliably and automatically identify the grant which provides the good data points.