Users report high CPU when diffractor starts. This is due to Diffractor building internal metadata indexes. 90% of the CPU is used by building a table of duplicates based on thumbnail visual hashes. This should be delayed until the user actually searches for duplicates based on image similarity.