Using the TYDL causality test, this paper attempts (i) to investigate the existence of shift contagion among a large spectrum of financial markets during recent stress and stress-free periods and (ii) to propose a new approach of portfolio management based on the minimization of the causal intensity. During the COVID-19 crisis period, the shift contagion analysis not only reveal a tripling of the causal links between the markets studied, but also a change in the causal structure. Beyond the initial impact of the COVID-19 crisis on financial markets, policy interventions seem to have helped in reassuring market participants that the further spread of financial stress would be mitigated. However, the Russian-Ukrainian conflict, and the high degree of uncertainty it entailed, has again exacerbated the interdependencies between financial markets. In terms of portfolio analysis, our minimum-causalintensity approach records a lower (respectively higher) reward-to-volatility ratio than the Markowitz (1952 & 1959) minimum-variance traditional approach during the pre-COVID-19 (respectively prewar) period. On the other hand, both approaches, the one we propose in this paper and the minimum-variance approach, record negative reward-to-volatility ratios during crisis periods.