Should a trader's backtest include periods of unusual volatility?

A strategy tested only during calm market conditions may behave differently during sharp price movements. Testing across a variety of environments can expose weaknesses that would otherwise remain hidden. How do you choose the historical periods used to test your strategy?