Originally Posted by
yoganmahew
Netflix do not care that two people watch the same movie at once.
My point is that they don't just let people develop software and hope it works when something crashes. They continually and deliberately kill random services in their system, so every piece of code
has to handle failures correctly; if it doesn't, they find out very quickly.
That means that, when something crashes unexpectedly, the rest of the system keeps going. Because it's been developed from the ground up to expect that to happen.
Whereas in software which isn't tested that way, I've seen something as simple as a debug message in code that's only called in a failure case kill the system because it accessed an invalid variable. That failure case was never seen in testing, so the code was never executed until it happened for real. And then it made the failure much, much worse than it would otherwise have been.