After 30+ years of business programming it's hard to pick one event.
But I definitely have to go with one concept:
The most important target audience for code is the maintenance programmer.
Can I tell one story? (Tough. I'm going to anyway.)
Many years ago, back in the days when a minute of CPU time cost about as much as an hour of programmer time (nowadays the ratio is more like a MONTH of processor-core time per hour of programmer time), I was working in a batch-Cobol shop and was asked to help with a maintance problem outside the projects I usually worked on.
The program in question, right in the middle of its work, did a complex calculation about electrical power management.
My relevant background for the task:
* I speak fluent, native (American) English. Amy, the programmer attempting to do the maintenance, only sort of spoke English.
* In college, I had dropped an electrical-engineering course.
* I had more programming experience in more languages, and a bit of a reputation for solving code puzzles.
So I looked at this mass of Cobol code that Amy had identified as the area where the bug apparently existed. Yep, it was the power-management calculation. After ten years, after the people who created it were gone, the client had realized it was calculating incorrectly.
Most of the program was unusually clear and comprehensible for Cobol. Excellent style, reasonably good technique. Nice meaningful variable names, but not absurdly long ones.
Then there was this part - about eighty lines. Amy could not make heads or tails of it.
Neither could I, for a couple days. Even after I noticed that the first third of the block was just moving data from variables with names like (making them up twenty years after the fact) Killowatt-Hours-Per-Day to other variables with Fortran-compliant names like FFGFXKCD (not to be confused with FFGFKXCD), and the last third was moving the data back.
I suggest:
1) Don't do this sort of calculation in Cobol.
2) If you're going to do it in Cobol, have a Cobol programmer write it. Not a Fortran programmer who's never seen Cobol before.
3) A programmer who understands the subject matter and has tried to maintain a program before, would be a nice touch too. I think the Fortran programmer was missing at least one of those attributes, but couldn't determine which one - possibly because I didn't understand the subject matter.
But after about four days I figured out what the formula actually being calculated was, identified a part that looked wrong to me (and in fact was wrong), and had Amy send it off to the client for feedback. Got the correct formula back a few days later, and replaced those eighty lines of cross-species monstrosity with ten lines of pure Cobol that Amy understood.
(I have no doubt that in a Fortran program written by a competent Fortran programmer with a reasonable understanding of the subject matter - the situation that really should have been in place, at that time, for this project - it would have been one or two lines.)