The RIAA 'case' against YouTube-dl is based on test code that downloads 10 kB

The interesting bit would be: are the 10 KB even persisted, or are they discarded after decode?

IANAL, but can well imagine this could have an impact on the interpetation of purpose.