From what I read and experienced, it makes the size of the Git repository much bigger than it should be with how Git tries to keep copies of old binary files (specifically the data and hash) since it cannot diff those unlike plaintext. You eventually have cache of the old files in the repo if you're not deliberate with setting --depth when cloning.

> From what I read and experienced, it makes the size of the Git repository much bigger than it should be...

Of course, it's a trade-off, but how is that particular one more painful than being at the whims of (and paying for) a third-party provider for something as mundane as static image files?