process scan results as soon as possible.
this should reduce the memory requirement when scanning because we do not keep all tags in memory anymore (unless all files are in the same directory, but the collection is probably not large in that case)