“If the object is not found in R2, it will simultaneously be returned from your S3 bucket and copied to R2.”
Sippy copies each object out of S3 the first time someone asks for it, on egress you were already paying. Data nobody reads never moves, and never runs up the one-time AWS transfer bill. Amazon built the lock-in on exit fees, and Cloudflare routes around them one request at a time. The catch is the usual one. You are swapping one vendor’s bucket for another’s.