A few years ago a friend and I were debugging a website he'd built that scraped data from Youtube, reorganized it, and built a web page that automatically categorized and assigned metadata to it. It was quite effective, and got more traffic than he expected, but it's value came in how it shed light on all sorts of bugs Youtube has.

One of the things we discovered is that Youtube actually runs from three content delivery networks who all have to agree on what data is to be displayed. The user data, such as your profile, watch history, and subscriptions, are in the first CDN. The video description, title, and comments attached to the video are in the second. The video itself and related data such as ad placement, categorization, and closed captions, are on a third CDN.

If the CDNs don't agree or one doesn't get updated in time before the page draws itself, you get mismatched descriptions/titles, mismatched comments, or some of the strangeness people have noticed where the timeline "watched" preview in the thumbnail is wrong by showing the user stopped watching earlier in the video than they actually did.

Sounds very interesting! (The website, and well also the bugs). Is the site still operational?

As far as I know, not any longer. At least publicly facing. The changes to Youtube in 2021 broke some of the functions, and he decided not to keep maintaining it and instead to move on to another project. The website was called EDM Engine, and was designed as a music marketing system that followed many of the design principles of the Vidya Internet Playlist created by the Aersia community. There is still a Facebook page that's up that might have screenshots of what it looked like, but I know people's aversion to going on that site.