This is a big understatement. :-)

Lead images (and all other images) are loaded asynchronously from the article text, so you can get reading while the app requests the images. Most people won't really notice this, but those on slower connections definitely will. Face detection runs on the lead images after it's downloaded, delaying its display, as we didn't want to readjust the image shortly after displaying it due to that being quite jarring. By reducing the size of the image used for face detection, we've also significantly reduced the amount of time it takes for faces to be detected. In some cases, the reduction was around an order of magnitude (500ms to 50ms). See the Phabricator task for more information.

So, you should all notice lead images appearing a fair bit faster than they did before! Hooray!

Also, the patch that implemented this was one of the first by our newest team member, Michael. Nice work, Michael!

Dan