Abstract <p>The central point of this work is the crystal size distribution (CSD) data on primitive olivine cumulates from two large layered intrusions located in Russia: the Yoko-Dovyren massif (Northern Baikal region) and Monchepluton (Murmansk region, Kola Peninsula). Dunites, troctolites, harzburgites, and chromitites have been studied. Ninety-one crystal size distributions of olivine and chromite, acquired mostly manually in thin sections, are presented. The CSD shape tends to be loglinear, bimodal (kinked), or lognormal. Number of CSDs for the selected zone or the part of stratigraphic cumulate succession in the intrusion is enough to reveal sustainably repeating CSD patterns, which provide a clear kinetic fingerprint of a some point in the dynamic evolution of crystallizing magma bodies and protocumulate crystal mushes. The CSDs kinetic record in turn may be partly deciphered by comparing the natural CSD forms with mathematical or qualitative models of distribution formation. We used previously published models for explanation of different CSD shapes and also proposed a modified model with nonequilibrium <i>T</i>-<i>t</i> path for the origin of the primary loglinear distribution widely known in magmatic rocks. The first idea of proposed modification is simultaneous accounting in a single model experimentally known bell-shaped dependence of nucleation and growth rates on supercooling and also experimentally proved dependence of the growth rate on the grain size (proportional growth). Previous models include this effect only separately. The second idea is the dependence of the intensity of the proportional growth effect also on supercooling. To this stage of calculation the original equation is proposed. The selected model for the natural CSD pattern formation is additionally verified by studying the growth and dissolution history of individual grains, which is recorded in chemical or morphological zoning of crystals. Phosphorus and other trace element zoning of olivine was analyzed using electron microprobe mapping. Contrast BSE images were used for the observation of zonal inclusion distribution in chromite. To take into account that the most common case is the open-system behavior of magma meaning that crystals move relative to the mother liquid during crystal population formation, the additional element is required to prove the dynamic process suggested by kinetics. This element is the study of bulk geochemistry and its correlation with CSD types, which reflects how the particular area of crystal mush with specific CSD differs from its previous state and how step by step it evolves away from the initial magma. Following this, CSD and zoning data for a number of olivine-rich rock samples were correlated with the bulk chemical composition of the rocks. The last aspect of the approach described in the paper involves assessing the thermodynamic state of the initial magma and crystalline mush of the studied region of a chamber at the time of local subsystem closure, i.e. geochemical thermometry. Here we used previously published or new estimates of the temperature of the parental magmas or mushes of studied rocks. Following the aforementioned way, a number of processes have been characterized and supposed. Those include: (1) magma mixing accompanied by olivine skeletal growth, (2) compaction of a crystal mush by pressure-solution, (3) infiltration of nonequilibrium porous melt through a crystal framework coupled with dissolution of accessory ore minerals or liquids (sulfide liquid and chromite), (4)&#xa0;chromite redistribution to overlying layers of crystal mush, (5) reverse peritectic reaction during decompression (for the Monchepluton magmas).</p>