NaN: the number that poisons every calculation it touches
NaN looks like a harmless floating-point oddity until it crosses a boundary. One invalid calculation can become a missing chart, a corrupt aggregate, a broken equality check, or a model training run that quietly collapses.