How a two-day workshop saved a year of wasted effort — and why killing a feature is sometimes the best outcome a QA can deliver.