I think the size and complexity of the program is the reason why we need to think first before we look at the code. I'll probably look at the obvious indicators (logs, stacktrace etc) and then the "thinking" could begin. Obviously with a huge program there are more probability for the local fix to happen because there are so many moving parts