[ericbarrett]

To satisfy the "third place" you might look into storing these files in Amazon S3 Glacier, which is about $1/TB/month, so long as you don't read them back. (Retrieval is delayed, batched, and expensive.)

It would take some engineering: I might store each optical disk image as a compressed image file, for example (zstd would be good for the large amount of data), to avoid metadata charges. Fun to think about.