Diversification: does it really fail, when you need it most?

Diversification has been labeled as the “only free lunch in investment management.” This conventional knowledge has been under attack for many years as investors observed that correlations of risky assets have been rising during periods of increasing systematic risks. Higher correlations are painful, but they usually do not happen ceteris paribus. In times of very volatile markets, even high correlations will proof to be useful and diversification benefits might increase rather than decrease. What are the conditions for this to happen?