Instead of thinking "same image on different domains, think "hidden uniquely-named single pixel image".

That’s the same thing. In order for that tracking method to work, this uniquely named pixel has to be loaded while visiting multiple sites. So it ends up being multiple domains referencing the same image from some tracker resource.

So a Facebook like button? Or a Twitter logo link?

I suspect you'll find this is _way_ more common than you expect.

(Also, if you thing Goog aren't doing this with their font CDN or various javascript library coaches, even (or especially) for sites without Google Analytics... I've got a bridge to sell you...)

It would be easy to check that...

It answers this question in the article; negligible performance hit

The most common example I could think of (other than trackers) would be aggregator sites. If the aggregator shows an image that was originally from a destination article or if comments link to a source for some content.

Good question. I'd guess that the chance of that happening is very small. But if that optimization exists maybe it's not that uncommon?