I also like the way the article is broken down into the backend, API, frontend and mobile. This isolated approach really highlights the different struggles each aspects of the product has, while dealing with what is essentially a shared concern: performance.
What I also found interesting is the fact that they were able to come up with a pretty accurate guess in terms of the expected traffic.
> "We experienced a maximum tile placement rate of almost 200/s. This was below our calculated maximum rate of 333/s (average of 100,000 users placing a tile every 5 minutes)."
Their guess ended up being a good amount above the actual maximum usage, but it was probably padded against the worst case scenario. The company that I work for consistently fails to come up with accurate guesses even with our very rigid user base, so it's pretty impressive that Reddit could accommodate the unpredictable user base that is the entire Reddit community.