Diamond-bearing gravels of the Lichtenburg-Ventersdorp area of the North West Province are associated with north-south orientated sinuous ‘runs’ that occur almost entirely on a flat erosional surface of the Malmani dolomites (Transvaal Supergroup) at some 1,500 m elevation. East to west, this dolomite plain measures 150 km, and north-south it is on average 40 km wide. This unconformity, which first developed before the Pretoria Group sedimentation over a period of at least 80 Myr, is marked by siliceous breccias (palaeo-karst infill) and conglomerates (reworked breccias). It was exhumed in pre-Karoo and post-Gondwana times. Glacial pavements and remnants of thin Lower Karoo sediments are also found on this polyphase surface. The gravels that make up these ‘runs’ and sinkholes directly or indirectly linked to these runs, are coarse-grained, very poorly-sorted, and are best described as diamictites. The ‘runs’ are narrow, elongated, generally positive ridges that meander across the dolomite surface and are up to 30 km long and between 80 to 300 m wide. They have always been regarded as post-Cretaceous drainage features linked to southward-flowing river systems. Diamonds were discovered in these ‘runs’ and they have produced some 12 million carats. However, no Cainozoic fossils or artefacts have ever been found in almost 90 years of mining. From new field evidence, geomorphological studies, age dating from inclusions in diamond and zircon and clay analyses, it is proposed that these coarse-grained runs represent proximal palaeo-eskers of the last deglaciation of the Dwyka continental ice sheet, that are preserved on this ancient ‘palimpsest’ surface. The age of the deposit is constrained by two populations of agate within the diamictites that are linked to two separate volcanic units of the Pretoria Group. In addition, the youngest crustal zircon ages from the gravels are 1 Ba, but mantle zircons from Lichtenburg suggest that these have been derived from Cambrian age kimberlites. Analysis of inclusions in diamond support a Neoproterozoic to Cambrian source for the diamonds, so the absence of diamonds from Mesozoic kimberlites and Cainozoic fossils within the gravels support the conclusion that the runs are of Karoo age.