In plutonic systems, magma mixing is often modelled by mass balance based on whole-rock geochemistry. However, magma mixing is a chaotic process and chemical equilibration is controlled by non-linear diffusive–advective processes unresolved by the study of bulk samples. Here we present textural observations, LA-(MC-)ICP-MS trace element and Sr–Nd isotopic data of accessory apatites and titanites from a hybrid granodiorite of the Neoarchean Matok pluton (South Africa), collected in a zone of conspicuous mixing between mafic and felsic magmas. Apatite grains mostly show a pronounced zoning in CL images, corresponding to abrupt changes in REE and HFSE concentrations recording their transfer through compositionally different melt domains during mixing. These grains crystallized early, at temperatures of 950–1000 °C. Titanite grains crystallized at temperatures of 820–900 °C (Zr-in-sphene thermometry). They show limited intra-grain chemical variations but huge inter-grain compositional scatter in REE and HFSE, pinpointing crystallization within a crystal mush, from isolated melt pockets having different composition from one another owing to incomplete chemical homogenization and variable Rayleigh fractionation. These chemical–textural characteristics, in combination with partitioning models and Polytopic Vector Analysis, point to “self-mixing” between co-genetic dioritic and granodioritic/granitic magmas. Both resulted from differentiation of mantle-derived mafic melts, showing that mixing does not necessarily involve magmas from contrasted (crust vs. mantle) sources. Systematic variations in εNdt (−4.5 to −2.5) and 87Sr/86Sr(i) (0.703–0.707) of titanite and apatite grains/domains crystallized from the two magmas point to an isotopically inhomogeneous mantle source, which is not resolved by bulk-rock isotopic data. Interaction between the two magmas must have occurred at relatively high temperatures (ca. 900°C) so that their viscosity contrast remained low, allowing efficient mechanical mixing. Despite this, chemical homogenization was incomplete, as recorded by diffusive fractionation between REE–HFSE and Sr. Modelling thereof reveals that chemical exchange between the liquid phases of the two mixed magmas did not last more than a few tens to hundreds of years. The chemical equilibration between mixed magmas thus strongly depends on the considered elements and observational length scales.