The term, from Nassim Taleb, has three parts and traders usually remember only the first. The event must be a surprise given the observer's information, it must carry severe consequences, and it must be explained away as obvious in hindsight - see hindsight-bias.
Most things labelled black swans are not. A sector crash after two years of obvious excess is a grey swan: foreseeable in kind if not in timing. Genuine black swans are rarer than the vocabulary suggests, and using the term for ordinary large losses is a way of avoiding responsibility for sizing.
The defence is structural, not predictive. You cannot forecast the event, so you build a book that survives one: capped worst-case-loss per position, no strategies whose loss is unbounded, liquidity kept in reserve, and leverage that assumes the model is wrong. Robustness costs return in ordinary conditions. That is the price.
Related: fat-tails, worst-case-loss, stress-testing, tail-risk