Play It Safe by Testing
I love rock climbing.
I love climbing, even though it frightens me. There's something thrilling about trusting your life to a bit of rope and a few metal pieces. I do it because I trust that the tools that I use have been built to withstand heavy weights, prolonged use, and long falls.
Take, for example, a carabiner. These small metal links need to meet some very specific criteria in terms of applied force they can withstand. Those criteria are much higher than what a normal user will ever need. Further, they are tested against all sorts of misuse. What if a carabiner is clipped backwards? What if it’s cross-loaded? What if the gate gets stuck on the rope, what if it's used for ice climbing, what if it’s thrown on rocks 200 times before it's used… An untested product could break and cause serious harm to the user, and consequently a lot of problems for the manufacturer.
I love video games.
See where I'm going with this?
Sure, at least in terms of physical harm, there's less exposure for a games studio if the game releases and the infrastructure breaks immediately. But the consequences can be financially devastating. Months, if not years in development, wiped away because the volume of players during the launch window prevents those very players from discovering your game. A couple days pass, those players move on to something else, and are unlikely to give your title a second chance. This is a costly mistake.
So, once you have tested your game in all sorts of controlled settings (playtesting, usability, functional QA, compliance QA…), you need to make sure that you give a few sessions to your game "in the wild". I'd say "one session", but I've yet to see a title go through a round of load testing and survive unscathed. Like clockwork, the cycle of using real-world scenarios is anticipation, surprise, and horror, and a few days or weeks later… another load test session with an improved product.