The grid of images seemed considerably large; my guess is if the page was just a static grid of images it would increase their bandwidth costs considerably, and might give rise to issues where if you scroll down far, it would still be loading images from much higher up. Using this technique it looks like when you scroll somewhere, it can load the images that you have navigated to. Maybe modern browsers are smarter in this respect? I'm not sure.
No, sadly modern browsers are not smarter. If you make a page with thousand of images then they will load all of them. There are JS libraries to fix it, but they make images inaccessible for bots.
