Backblaze should be able to handle 17TB of data.
Not sure what happened in this particular case as I'm not familiar with it, but I'd venture to guess that something occurred on the computer that caused the client to think it needed to re-upload those 13TB of data, so it scheduled them for re-upload.
However, I'm not sure how many of their clients have single drives with 17TB of data. Perhaps this is an edge case that they didn't adequately test for.
And since a vast majority of their customers will not have more than 3TB on any single drive, I can see why this might not be well tested.