Jep you're right. It's doing something with 64k JFIF application segments ... wtf.

Well, it's still using the trick I pointed out, placing the ZIP file index at the end.

So from one viewpoint it's a JFIF("jpg") file with large application segments containing the zipfile data for the shakespeare.part0xx.rar files.

From another viewpoint, from the back of the file. It's a incremental zip file (not compressed in one go), with the garbage bytes (the "overwritten") bytes in the zipfile updates forming a valid JFIF file.