Hm, that level of specific backend detail I'll have to refer to Development for an answer, but in a nutshell, our download links for remotely-hosted files work like this:
The link URL we issue to the buyer looks like http://www.e-junkie.com/[...etc.] When the buyer clicks their link to that URL on our server, our backend looks up what product it goes to, and if a Remote Product File URL had been specified for the product, we stream a copy of the file from your file's static URL, through our Web cache, to the buyer, so insofar as the buyer's end can tell, the file is coming from us.
Once this initial download has completed successfully, we retain our cached copy of that download in our Amazon S3 provisions; for every subsequent download, we first check to see if the original copy at your URL has changed, and if not, we simply serve our cached copy from S3 to the buyer. This eliminates unnecessary transfer-bandwidth surcharges or limits for both you and us, and ensures the quickest and most trouble-free download possible for the recipient.