Law of the untested part
The part of the program that breaks in production is always the one nobody thought of testing, because it was too simple to get wrong.
Tests always focus on the difficult parts: complex calculations, algorithms, payments. Fortresses are built around those. Then the website collapses because of the 'Contact us' button, which nobody had ever clicked because 'it's only a form'.
It's the computing version of the medieval castle with walls three metres thick and the back door tied shut with string. The enemy doesn't scale the walls. The enemy, quite simply, knocks round the back.
The high priest of quality has been repeating the same sermon for years: 'Whatever isn't tested is broken, you just don't know it yet.' It's a frightening sentence, because it's true. Now I test the trivial things too, one by one, with the zeal of a solicitor. The client asks why it takes me so long. I tell them I'm checking the back door.