Like Steve and Simon I have developed systems that ran for decades without revision and without error other then data or other operational errors. That does not mean that I am an ace programmer, only that every once in a while we got it right, and it stayed right.
I too have developed code that works for years without modification. But I'm not going to project those small successes to all my code or all the possible code.
There are systems and processes that can be defined and maintained simply. But there's a lot of what physic guys call the "n-body problem". We've dealt with situations where n=2 and our code can run forever for a given situation. But with the combined complexity of hardware, operating systems, RDBMS, programming languages, requirements, users, networks, malware, etc, n is more likely greater than 3.