Perhaps the logic is that if you don't use the data for a month and a half and your only copy of the data is in ubuntu one it must not be that important?
Do you really want to wait for 5 hours just to figure out you've requested a wrong folder and the photos you're looking for were shot in June, not July?
I would assume that such a service would generate metadata indexes (incl. thumbnails) that are stored on more accessible media. That is, in fact, one of the main reasons a technical person can't just use Amazon Glacier for this directly, in the same way they could use S3 directly for Dropbox's use-case.
