> The only missing bit would be the ability to replicate production traffic, as some bugs only happen under very high traffic by a large number of concurrent u

> The only missing bit would be the ability to replicate production traffic, as some bugs only happen under very high traffic by a large number of concurrent users. This is however at best difficult or nearly impossible to do. Not sure I'm missing something here. Surely you could sample some prod traffic and then replay it with one of the many load test tools out there. You might lose in the geographical distribution, but load testing a web server with 2k TPS sounds a bit trivial.