Mineral and whole rock elemental compositions used for initial classification according to the scheme of Aulbach & Jacob (2016). Left panels show samples in this study and other studies from the Kaapvaal craton (Carswell et al., 1981; McCandless & Gurney, 1989; Caporuscio & Smyth, 1990; O'Reilly & Griffin, 1995; Viljoen, 1995; Jacob et al., 2003; Viljoen et al., 2005; Jacob et al., 2009; Gréau et al., 2011; Huang et al., 2012; Aulbach & Viljoen, 2015; Shu et al., 2016; Radu et al., 2019; Aulbach et al., 2020a, 2020b; Hardman et al., 2021); corundum (crd)-bearing and corundum-diamond (dia) bearing samples are shown with separate symbol; right panels show corundum-bearing and diamondiferous (dia) eclogites only, plus inclusions in diamond from the Kaapvaal craton separately for clarity (Moore & Gurney, 1985; Moore et al., 1991; Appleyard et al., 2004; Viljoen et al., 2010, 2018; UCT database 2018 as cited in Stachel, 2021). a. and b. Grossular component in garnet expressed as Ca# (molar Ca/(Ca + Mg + Fe_total + Mn)) vs. jadeite component in clinopyroxene allowing to distinguish true eclogite from pyroxenite xenoliths; c. and d. ∑HREE vs. Eu/Eu^* in reconstructed bulk rocks allowing to distinguish gabbroic eclogites and pyroxenites (with Eu/Eu^* >1.05, indicative of substantial plagioclase accumulation, therefore a cumulate protolith) from non-gabbroic varieties that may have originated from melt-like protoliths (if not later strongly modified by mantle-derived melts, which can mute Eu anomalies; Aulbach et al., 2020b), and e. and f. Ca# vs. Mg# (Mg/(Mg + Fe_total) molar) in garnet allowing to distinguish high-Mg, low-Mg and high-Ca eclogites (Aulbach & Jacob, 2016; Aulbach et al., 2020a, 2020b). Shown for comparison in d. are oceanic crustal sections comprising intrusive and extrusive rocks from various settings/localities (Klein et al., 1991; Zimmer et al., 1995; Benoit et al., 1996; Hart et al., 1999; Godard et al., 2009).