There is no consensus on the processes responsible for near-coeval formation of Archaean continental crust (dominantly tonalite–trondhjemite–granodiorite: TTG), greenstone belts dominated by komatiitic to tholeiitic lavas (KT), and sub-continental lithospheric mantle (SCLM). The Douglas Harbour domain (2.7–2.9 Ga) of the Minto Block, northeastern Superior Province, has two TTG suites, the western and eastern Faribault–Thury (WFT and EFT), with embedded KT greenstones. Tonalites of both suites have high light/heavy rare-earth element ratios (L/HREE), high large ion lithophile element (U–Th–Rb–Cs–La: LILE) contents, positive Sr–Pb anomalies, and negative Nb–Ta–Ti anomalies. Such typical Archaean TTG signatures are commonly explained by melting of subducted oceanic crust, but could also originate by melting the base of thick basaltic plateaux formed above mantle upwellings (plumes), leaving behind restites containing pyroxene, garnet, and rutile. Field relationships (in situ segregation veins), phase equilibria (hornblende stabilized at lower crustal pressure), petrography (corroded epidote and muscovite phenocrysts, rare plagioclase phenocrysts), and trace element models, all imply that FT tonalite to trondhjemite evolution reflects hornblende-dominated fractional crystallization, not partial melting of subducted crust. The geochemistry of parental FT tonalites can be modeled by 15–30% melting of FT tholeiitic metabasalts, with residues of eclogite, garnet-websterite, or hornblende-garnet websterite. A minor residual Ti-phase such as rutile is also needed to generate negative Ti–Nb–Ta troughs in the TTGs. However, large volumes of eclogitic restites complementary to TTG are not observed either at the base of Archaean crustal sections, or in the SCLM. Additional problems with slab-melting models include: (a) the rarity of lithologies and associations characteristic of active margins (ophiolites, andesites, blueschists, accretionary mélanges, molasse, flysch, high-pressure belts, and thrust-and-fold belts); (b) the need to deliver plume-derived KT melt through the slab; and (c) extracting enough TTG melt from a subducting slab in the time available (200–300 my). In the plateau-melting model, heat for crustal anatexis is supplied by ongoing KT magma derived from mantle upwellings. However, SCLM rocks differ from predicted 1-stage mantle melting residua; and the voluminous residual eclogites complementary to TTG generation somehow need to be removed. These two problems might solve one another if the dense crustal restites disaggregated and mixed into the underlying depleted mantle. Mantle melting slows upon exhaustion of Ca–Al-rich phases, with large temperature increases needed to extract more melt from harzburgite residua. Physical addition of delaminated crustal restites would refertilize the refractory mantle, allowing extraction of additional melt increments, and might explain the ultra-depleted and orthopyroxene-rich nature of the SCLM. A hybrid source composed of 10% eclogitic restite of EFT tonalite generation, mixed with harzburgitic residues from 25% melting of primitive mantle, yields model melts with trace element signatures resembling typical Munro komatiites. Variations in the mineralogy and geochemistry of the delaminated component might account for the diversity of komatiite types. Degassing of hornblende-rich delaminated restites would transfer LILE to surrounding depleted mantle and could generate boninites. Fusion of undepleted metabasalt sandwiched among denser restites could generate sanukitoids. Mantle melt pulses generated by catastrophic delamination events would underplate nascent TTG crust and trigger renewed crustal melting, followed by delamination of newly formed eclogitic restites, triggering additional mantle melting, and so on. I posit that delamination of crustal restites catalyzed multi-stage melting of the SCLM and maturation of the Archaean continental crust. Thus, Archaean crust and SCLM are genetically inter-linked, and both form above major mantle upwellings.