I don't think anything has ever surpassed the Stonesurfers Pyroto question database, even though that's quite some time ago.
Some background: Pyroto was a weird sort of social trivia game. There was a large but finite database of trivia questions, and it you played enough you would see repeats. But each player had a limited number of turns per day, and a limited time to answer each question - one try, right or wrong. So it led to players saving the questions, and the answers they tried, right or wrong.
Now, as I said, it was a social trivia game. There were official in-game alliances of a sort, and this led to external-to-the-game teams who shared their saved questions and answers.
Our team had a Yahoo group, and people started entering their questions and answers into the database which the Yahoo group provided. This worked, for a while. Eventually, though, we exceeded the fixed 1000 row limit that Yahoo put on the database they provided. Also, this wasn't maintainable. There were lots of duplicate and near-duplicate questions due to people re-entering things we had already found.
So another system was layered over this. Members used a web email form to submit their questions and answers in a format that a program I wrote would read. This program would consolidate duplicates, check based on some heuristics I wrote for near-duplicates that maybe were really the same question, and when it was all done, wipe out the database we had last used, and bulk-upload the new version into the database.
This worked for a while. But the Yahoo group database had a hard limit of 1000 rows, and once we got this operation going, pretty quickly we exceeded this. We were allowed 10 of these databases, though, so at first we split them up by level. There were 6 or 7 different levels which were exponential; you'd play twice as much at each level compared to the preceding one, and the number of questions in each level was roughly proportional to the time you'd spend there.
Naturally, we pretty quickly exceeded the 1000 rows on just the highest difficulty level, so I just started splitting up those high difficulty levels into multiple databases, and consolidating the easy ones with few questions into a single database.
Eventually, we had over 9000 questions and the writing was on the wall. I scrapped the idea of using the Yahoo database ONLY AT THIS POINT and instead wrote a search engine using a local Python web server each player would run queries against, which made it much easier to find answers - the same similarity code I was using to find near-duplicates was then available for everybody all the time, and the question updating was built into the thing too. But the only reason I made the real solution was because the lame thing we had been doing had hard limits built in!