We devise a simple yet versatile strategy to perform trend-cycle decompositions in severe crisis periods, such as the COVID-19 pandemic period. The proposed strategy propels a great deal of volatility during this period into pandemic-specific shocks, with minimal impacts on non-pandemic disturbances. We start by estimating two unobserved components models until 2019:4, for Portugal and the euro area. We then introduce several pandemicspecific disturbances and estimate their variances during the 2020-21 period, keeping fixed all remaining model parameters. Finally, we bring together the information from both estimation stages through a piecewise linear Kalman filter, assuming such heteroskedastic environment. Our strategy has the attractiveness of generating negligible historical revisions when the 2020-2021 period is added to the estimation sample, despite the large pandemic disruption. Results suggest that innovations affecting the cycle are key drivers of GDP during the pandemic period, while yielding negligible historical revisions.