Repair an inconsistent alloy equilibrium assessment
A materials-thermodynamics group is replaying literature-linked alloy studies with the supplied historical pycalphad source and databases. The calculations complete and return finite datasets, but several observations no longer agree with the group's archived results:
- a composition-temperature phase boundary moves when an equivalent database representation is admitted to the same assessment;
- the associated phase fractions and internal-state evolution cease to form a physically consistent transition;
- chemical-potential and heat-response curves disagree with the corresponding baseline calculation; and
- the equilibrium topology depends on a representation that should not create a new thermodynamic state for the requested components.
Run the offline workflow, inspect the databases and source snapshot, and repair the scientific implementation. Determine the common cause rather than treating the symptoms independently. The repair must generalize to other valid CALPHAD database representations and supported phase states while preserving public APIs and unrelated model behavior.
Do not hard-code the supplied grids or generated artifacts, and do not modify the scientific input files. The public workflow only establishes that the real project calculation runs and saves results; correctness is evaluated on independent scientific cases.