Pentlandite is the dominant Ni-hosting ore mineral in most magmatic sulfide deposits and has conventionally been interpreted as being entirely generated by solid-state exsolution from the high-temperature monosulfide solid solution (MSS) (Fe,Ni)1–xS. This process gives rise to the development of loops of pentlandite surrounding pyrrhotite grains. Recently it has been recognized that not all pentlandite forms by exsolution. Some may form as the result of peritectic reaction between early formed MSS and residual Ni-Cu–rich sulfide liquid during differentiation of the sulfide melt, such that at least some loop textures may be genuinely magmatic in origin. Testing this hypothesis involved microbeam X-ray fluorescence mapping to image pentlandite-pyrrhotite-chalcopyrite intergrowths from a range of different deposits. These deposits exemplify slowly cooled magmatic environments (Nova, Western Australia; Sudbury, Canada), globular ores from shallow-level intrusions (Norilsk, Siberia), extrusive komatiite-hosted ores from low and high metamorphic-grade terranes, and a number of other deposits. Our approach was complemented by laser ablation-inductively coupled plasma-mass spectrometry analysis of palladium in varying textural types of pentlandite within these deposits. Pentlandite forming coarse granular aggregates, together with loop-textured pentlandite where chalcopyrite also forms part of the loop framework, consistently has the highest Pd content compared with pentlandite clearly exsolved as lamellae from MSS or pyrrhotite. This is consistent with much of granular and loop pentlandite being formed by peritectic reaction between Pd-rich residual sulfide liquid and early crystallized MSS, rather than forming entirely by subsolidus grain boundary exsolution from MSS, as has hitherto been assumed. The wide range of Pd contents in pentlandite in individual samples reflects a continuum of processes between peritectic reaction and grain boundary exsolution. Textures in metamorphically recrystallized ores are distinctly different from loop-textured ores, implying that loop textures cannot be regenerated (except in special circumstances) by metamorphic recrystallization of original magmatic-textured ores. The presence of loop textures can therefore be taken as evidence of a lack of penetrative deformation and remobilization at submagmatic temperatures, a conclusion of particular significance to the interpretation of the Nova deposit as having formed synchronously with the peak of regional deformation at temperatures within the sulfide melting range.