New trace-element data of rutile in kimberlite-borne ~1.85 Ga eclogite and pyroxenite xenoliths from the central Slave craton, as well as ~110 Ma MARID xenoliths from the Kaapvaal craton, provide constraints on the origins of lithospheric and sublithospheric mantle variability in high field strength element ratios. Rutiles in eclogites and pyroxenites have Zr/Hf ranging from 20 to 62 and Nb/Ta ranging from 10 to 40. Rutiles in MARID xenoliths have Zr/Hf from 24 to 33 and Nb/Ta from 10 to 41. Calculated whole-rock Zr/Hf is suprachondritic for eclogites with suggested gabbroic protoliths and subchondritic for boninite-like eclogites; the latter is consistent with cpx-controlled depletion in the protolith source. Within each eclogite type, positive correlations of Zr/Hf with La/Lu and negative correlations with Lu/Hf likely reflect fractionation of cpx and/or plagioclase during crystallisation of the protoliths. Zr/Hf-Nb/Ta relationships of some MARID-type rocks, which are products of lithospheric mantle metasomatism, and eclogite xenoliths plot on a silicate differentiation trend, whereas other samples have higher Nb/Ta at a given Zr/Hf. Fractionation of a few percent rutile from an HFSE-rich mafic melt can generate a trend towards strongly increased Nb/Ta at minimally changed Zr/Hf in the residual melt. Superposition of rutile fractionation on the effects of silicate differentiation, which fractionates Zr/Hf more strongly than Nb/Ta, can explain the Zr/Hf-Nb/Ta relationships of most eclogites from the central Slave craton as well as those of MARID rocks, metasomatised peridotites and group II kimberlites. By contrast, Zr/Hf-Nb/Ta relationships suggest that Group I kimberlites are mixtures between depleted peridotite and carbonatite. Thus, high Nb/Ta is a signature of lithospheric processes and may not be important in deeply subducted eclogites that bypass extended residence in the lithosphere. Conversely, considerable primary Zr/Hf variability was inherited by the eclogites, which is indicative of the compositional diversity of ancient subducted oceanic crust, which is expected to have generated substantial heterogeneity in sublithospheric basalt sources.