What is this black magic?
It’s jpeg image with html content in comment section, which is not encoded in any way. When you try to load this in img tag, it’s normal image (comment are ignored) and displayed.
Browsers are trying really hard to make some sense of content (even with lots of bugs) for historical and compatibility reasons. In this case when they find tags, the parse them and show content. The rest of it (jpeg headers at the beginning, actual img content after comment section) is still there, hidden with css. Starting comment also ends with <!– to comment out image data.
Check the source and then developer console to see what browser actually thinks about the content.