In kesterite CuZnSn(S,Se) (CZTSSe) solar cell research, an asymmetric crystallization profile is often obtained after annealing, resulting in a bilayered - or double-layered - CZTSSe absorber. So far, only segregated pieces of research exist to characterize the appearance of this double layer, its formation dynamics, and its effect on the performances of devices. In this work, we review the existing research on double-layered kesterites and evaluate the different mechanisms proposed. Using a cosputtering-based approach, we show that the two layers can differ significantly in morphology, composition, and optoelectronic properties and complement the results with a large statistical data set of over 850 individual CZTS solar cells. By reducing the absorber thickness from above 1000 to 300 nm, we show that the double-layer segregation is alleviated. In turn, we see a progressive improvement in the device performance for lower thickness, which alone would be inconsistent with the well-known case of ultrathin CIGS solar cells. We therefore attribute the improvements to the reduced double-layer occurrence and find that the double layer limits the efficiency of our devices to below 7%. By comparing the results with CZTS grown on monocrystalline Si substrates, without a native Na supply, we show that the alkali metal supply does not determine the double-layer formation but merely reduces the threshold for its occurrence. Instead, we propose that the main formation mechanism is the early migration of Cu to the surface during annealing and formation of CuS phases in a self-regulating process akin to the Kirkendall effect. Finally, we comment on the generality of the mechanism proposed by comparing our results to other synthesis routes, including our own in-house results from solution processing and pulsed laser deposition of sulfide- and oxide-based targets. We find that although the double-layer occurrence largely depends on the kesterite synthesis route, the common factors determining the double-layer occurrence appear to be the presence of metallic Cu and/or a chalcogen deficiency in the precursor matrix. We suggest that understanding the limitations imposed by the double-layer dynamics could prove useful to pave the way for breaking the 13% efficiency barrier for this technology.