In a data lakehouse architecture; what does separating storage from compute primarily enable?

Decoupling storage (typically object storage) from compute (query engines) allows each to scale independently; reducing cost by paying only for what is used in each tier. The incorrect options ("Faster ETL pipelines by co-locating data with processing nodes", "Elimination of the need for a metadata catalog", "Automatic conversion of all data files to a single optimized format") are distractors that don't fully capture the concept described.