Gruta do Caldeirão features a c. 6 m-thick archaeological stratification capped by Holocene layers ABC-D and Ea, which overlie layer Eb, a deposit of Magdalenian age that underwent significant disturbance, intrusion, and component mixing caused by funerary use of the cave during the Early Neolithic. Here, we provide an updated overview of the stratigraphy and archaeological content of the underlying Pleistocene succession, whose chronology we refine using radiocarbon and single-grain optically stimulated luminescence dating. We find a high degree of stratigraphic integrity. Dating anomalies exist in association with the succession’s two major discontinuities: between layer Eb and Upper Solutrean layer Fa, and between Early Upper Palaeolithic layer K and Middle Palaeolithic layer L. Mostly, the anomalies consist of older-than-expected radiocarbon ages and can be explained by bioturbation and palimpsest-forming sedimentation hiatuses. Combined with palaeoenvironmental inferences derived from magnetic susceptibility analyses, the dating shows that sedimentation rates varied in tandem with the oscillations in global climate revealed by the Greenland oxygen isotope record. A steep increase in sedimentation rate is observed through the Last Glacial Maximum, resulting in a c. 1.5 m-thick accumulation containing conspicuous remains of occupation by people of the Solutrean technocomplex, whose traditional subdivision is corroborated: the index fossils appear in the expected stratigraphic order; the diagnostics of the Protosolutrean and the Lower Solutrean predate 24,000 years ago; and the constraints on the Upper Solutrean place it after Greenland Interstadial 2.2. (23,220–23,340 years ago). Human usage of the site during the Early Upper and the Middle Palaeolithic is episodic and low-intensity: stone tools are few, and the faunal remains relate to carnivore activity. The Middle Palaeolithic is found to persist beyond 39,000 years ago, at least three millennia longer than in the Franco-Cantabrian region. This conclusion is upheld by Bayesian modelling and stands even if the radiocarbon ages for the Middle Palaeolithic levels are removed from consideration (on account of observed inversions and the method’s potential for underestimation when used close to its limit of applicability). A number of localities in Spain and Portugal reveal a similar persistence pattern. The key evidence comes from high-resolution fluviatile contexts spared by the site formation issues that our study of Caldeirão brings to light—palimpsest formation, post-depositional disturbance, and erosion. These processes. are ubiquitous in the cave and rock-shelter sites of Iberia, reflecting the impact on karst archives of the variation in climate and environments that occurred through the Upper Pleistocene, and especially at two key points in time: between 37,000 and 42,000 years ago, and after the Last Glacial Maximum. Such empirical difficulties go a long way towards explaining the controversies surrounding the associated cultural transitions: from the Middle to the Upper Palaeolithic, and from the Solutrean to the Magdalenian. Alongside potential dating error caused by incomplete decontamination, proper consideration of sample association issues is required if we are ever to fully understand what happened with the human settlement of Iberia during these critical intervals, and especially so with regards to the fate of Iberia’s last Neandertal populations.