Context. Feedback by massive stars shapes the interstellar medium and is thought to influence subsequent star formation. The details of this process are under debate. Aims. We exploited observational constraints on stars, gas, and nucleosynthesis ashes for the closest region with recent massive-star formation, Scorpius–Centaurus OB2, and combined them with three-dimensional (3D) hydrodynamical simulations in order to address the physics and history of the Scorpius–Centaurus superbubble. Methods. We used published cold gas observations of continuum and molecular lines from Planck, Herschel, and APEX. We analysed the Galactic All Sky Survey (GASS) to investigate shell structures in atomic hydrogen, and used Hipparcos and Gaia data in combination with interstellar absorption against stars to obtain new constraints for the distance to the Hi features. Hot gas is traced in soft X-rays via the ROSAT all sky survey. Nucleosynthesis ejecta from massive stars were traced with new INTEGRAL spectrometer observations via ²⁶Al radioactivity. We also performed 3D hydrodynamical simulations for the Sco–Cen superbubble. Results. Soft X-rays and a now more significant detection of ²⁶Al confirm recent (≈1 Myr ago) input of mass, energy, and nucleosynthesis ejecta, likely from a supernova in the Upper Scorpius (USco) subgroup. We confirm a large supershell around the entire OB association and perform a 3D hydrodynamics simulation with a conservative massive star population that reproduces the morphology of the superbubble. High-resolution GASS observations reveal a nested, filamentary supershell. The filaments are possibly related to the Vishniac clumping instability, but molecular gas (Lupus I) is only present where the shell coincides with the connecting line between the subgroups of the OB association, suggesting a connection to the cloud, probably an elongated sheet, out of which the OB association formed. Stars have formed sequentially in the subgroups of the OB association and currently form in Lupus I. To investigate the impact of massive star feedback on extended clouds, we simulate the interaction of a turbulent cloud with the hot, pressurised gas in a superbubble. The hot gas fills the tenuous regions of the cloud and compresses the denser parts. Stars formed in these dense clumps would have distinct spatial and kinematic distributions. Conclusions. The combined results from observations and simulations are consistent with a scenario where dense gas was initially distributed in a band elongated in the direction now occupied by the OB association. Superbubbles powered by massive stars would then repeatedly break out of the elongated parent cloud, and surround and squash the denser parts of the gas sheet and thus induce more star formation. The expected spatial and kinematic distribution of stars is consistent with observations of Sco–Cen. The scenario might apply to many similar regions in the Galaxy and also to active galactic nucleus (AGN)-related superbubbles.