In reading that, my first thought was "why in the world would Uber do that?"
In every single performance tuning a scale story that I've read over the past decade, the very first point of order is: remove joins from high traffic queries. It seems like Uber has gone the complete opposite direction.
It's unrealistic to remove all joins (usually), and most joins aren't very expensive. In normal cases a join is strictly faster than running an extra query to get that extra data.
Joins start getting particularly tricky when you do expensive stuff like filter over the correlated results of many tables. But in those cases, there's also no easy alternative: you'll need to redesign the way you store your data, if possible.
But that is largely incorrect advice, given based on the fact that mysql never implemented any reasonable join algorithms, just nested loops. It didn't apply to real databases.
In every single performance tuning a scale story that I've read over the past decade, the very first point of order is: remove joins from high traffic queries. It seems like Uber has gone the complete opposite direction.