That’s a really good post with good & practical advice for rate limiting & load shedding. Straight away set me thinking about where in our system we could most benefit from introducing the various limiters.
One thing that struck me as odd was my understanding that it was the same system that handled test requests and non-test requests. Did that strike anyone else as odd? I think I’ve only ever worked on systems were test requests went to a totally different deployment, with different URLs etc.