Fail Safe. Shit is going to go wrong at the the most in opportune time.
Ensure your production environment will kill your applications at random times. At least you'll be practiced. Uptime is your enemy and gives you a false sense of security. Given enough time, eventually someone will trip over a power cord despite your triply redundant power supplies and backup systems.
I learned this when a memory leak became impossible to track down and so we just killed the process when it reached a certain point. That program was the most reliable thing I've ever seen, eventually every inopportune moment it could have died had been found and the code knew what to do. It would never crash despite always crashing.
Ensure your production environment will kill your applications at random times. At least you'll be practiced. Uptime is your enemy and gives you a false sense of security. Given enough time, eventually someone will trip over a power cord despite your triply redundant power supplies and backup systems.
I learned this when a memory leak became impossible to track down and so we just killed the process when it reached a certain point. That program was the most reliable thing I've ever seen, eventually every inopportune moment it could have died had been found and the code knew what to do. It would never crash despite always crashing.