•Thermodynamic analysis of UHP fluid inclusion chemistry and phase equilibria.•Marbles release solute-bearing COHS fluids during UHP Alpine subduction.•Post-entrapment chemical evolution of UHP inclusions is reconstructed.•Electrolytic fluid models are consistent with the inferred evolution. Subduction fluids play a crucial role in regulating long-term chemical cycles. Their characterisation is essential to understand the processes responsible for metasomatism, oxidation and melting of the mantle wedge. Both direct (fluid inclusion studies) and indirect (thermodynamic modelling) approaches to study subduction fluids have reliability issues due to the complexity of the investigated processes. Post-entrapment processes (e.g., solvent loss by diffusion or decrepitation and/or chemical reactions between host mineral and trapped fluid) are likely to modify the chemical fingerprint of ultra-high pressure (UHP) fluid inclusions, while thermodynamic modelling of solute-bearing fluids at UHP conditions is still at the beginning of its application. In this work, we apply and compare data obtained by both approaches for fluid inclusions trapped within UHP clinopyroxene from a chemically simple Ol-Cpx-Dol-Cal marble (Brossasco-Isasca Unit, Dora-Maira Massif, Western Italian Alps). Classical molecular-fluid thermodynamics is adequate to qualitatively describe the post-entrapment reactions between fluid inclusions and host clinopyroxene. However, an electrolytic fluid model is necessary to describe the chemical composition of the solute-bearing aqueous fluids at the peak metamorphic condition (H2O: 96.30 mol%/88.49 wt%; solutes: 3.61 mol%/11.34 wt%/2.08 mol/kg; other volatiles: 0.09 mol%/0.17 wt%) generated by progressive rock dissolution. Comparison of the model fluid composition with that inferred from the analysis of fluid inclusions clarifies the types and the extent of post-trapping chemical modification of the UHP fluid inclusions. Our data reveal that the fluid-host reactions carry up to 42 mol% of host clinopyroxene component in the fluid inclusion bulk composition, whereas the fluid inclusion decrepitation and the water diffusion in the host clinopyroxene (through dislocations and/or micro-fractures) cause an H2O loss ranging from 18 mol% to 99 mol%. Applying these approaches, we demonstrate that the most relevant post-entrapment process is H2O loss. We also demonstrate that some fluid inclusions did not experience post-entrapment fluid-host modification and, thus, preserve the original fluid geochemistry.