Right now when clicking to download the .fits of an image in a datalab session, it just uses the stored s3 url. This presigned url expires (I think in 24 hours after it was generated?), so that download will fail if your image was added to the session longer ago than that. What we should do instead if when download fits is called, it should call the archive /frames/?basename_exact={basename} to get a regenerated presigned url to download the .fits file from. Or it could call the backend which then calls the archive, either way, it needs to get a fresh presigned s3 url to be able to download the .fits file.