Abstract We study a model of rapidly cooling shocked stellar winds in young massive clusters and estimate the circumstances under which secondary star formation, out of the reinserted winds from a first stellar generation (1G), is possible. We have used two implementations of the model: a highly idealized, computationally inexpensive, spherically symmetric semi-analytic model, and a complex, three-dimensional radiation-hydrodynamic, simulation; they are in a good mutual agreement. The results confirm our previous findings that, in a cluster with 1G mass 107 M ⊙ and half-mass–radius 2.38 pc, the shocked stellar winds become thermally unstable, collapse into dense gaseous structures that partially accumulate inside the cluster, self-shield against ionizing stellar radiation, and form the second generation (2G) of stars. We have used the semi-analytic model to explore a subset of the parameter space covering a wide range of the observationally poorly constrained parameters: the heating efficiency, η he, and the mass loading, η ml. The results show that the fraction of the 1G stellar winds accumulating inside the cluster can be larger than 50% if η he ≲ 10%, which is suggested by the observations. Furthermore, for low η he, the model provides a self-consistent mechanism predicting 2G stars forming only in the central zones of the cluster. Finally, we have calculated the accumulated warm gas emission in the H30α recombination line, analyzed its velocity profile, and estimated its intensity for super star clusters in interacting galaxies NGC4038/9 (Antennae) showing that the warm gas should be detectable with ALMA.