First read from each edge would be two reads. One to get the data from the central S3 repo, one to serve to the end user. But, that's not double... after the fi

First read from each edge would be two reads. One to get the data from the central S3 repo, one to serve to the end user. But, that's not double... after the first read for each edge the additional read is no longer needed.