But second, they claim they've been doing this to images uploaded to Dropbox. [...] But what happens when they find out their new algorithm -- which compresses AND decompresses! -- has a bug when it hits a certain data condition, and sorry, all your images are corrupted because the EXIF data common to them all triggered the bug.
Assume that the engineers behind this aren't morons. Failing that, read the article. For every newly compressed image, Dropbox does a decompression and a bit-for-bit comparison with the original before replacing the original. If there's an image that triggers a bug that corrupts the image for whatever reason, their test will catch it before the original image is replaced.