Allende spinels display either direct spinel-melilite contact or an intervening boundary clinopyroxene between the two phases. Spinel-melilite interfacial regions in a Leoville B1 are more complex, with boundary clinopyroxene, as observed in Allende, but also variable amounts of glass, secondary calcite, perovskite, and an Mg-, Al-, OH-rich and Ca-, Si-poor crystalline phase that may be a layered double hydrate. One possible scenario of formation for the glass veins is that open system alteration of melilite produced a porous, hydrated aggregate of Mg-carpholite or sudoite + aluminous diopside that was shock melted and quenched to a glass. The hydrated crystalline phase we observed may have been a shocked remnant of the precursor phase assemblage, but is more likely to have formed later by alteration of the glass. In the mantle, boundary clinopyroxenes may have been crystallized from Ti-rich liquids formed by the direct dissolution of perovskite and an associated Sc-Zr-rich phase or as a reaction product between dissolving perovskite and liquid. In the core, any perovskite and associated Ti-enriched liquids that may have originally been present disappeared before the growth of boundary clinopyroxene, and the observed boundary clinopyroxene may have nucleated and grown from the liquid, along with the larger core clinopyroxene.