The problem is not that they suddenly hit this performance drop and solved it in some reasonable timeframe, or it was some rare occurrence affecting .01% of users

The problem is that it’s existed for 6 years, affected anyone and everyone, and would have been caught and resolved by basically anyone who attempted to look into the issue with fairly low effort/experience-needed. That is, the problem is that no one bothered to look into it, despite good reason to do so.