This is 100% correct. We also generated static snapshots of the origin and mirrored to S3. In the instance of something going awry with our primary origin servers, we could fail back to S3 as an origin.

EDIT: I'll add that our origin offload for WWW hovered around 98%.