1σ confidence contours for cosmological and nuisance parameters, as in Fig. 15, now for all forecasts conducted using synthetic stage-IV-like galaxy samples (Sects. 3.2.1 & 3.2.2), including: the true redshift distribution $n_z^{\mathrm{true}}$ with the ‘do nothing’ model (i.e. the idealized case; solid contours); the shifted distribution $n_z^{\mathrm{shift}}$ with the ‘do nothing’ model (the uncorrected case; dotted contours); and the shifted distributions with the shift δz_i model (the corrected case; dashed contours); each performed for the γγ (red), 3 × 2 pt (blue), 3^ph × 2 pt (green), and 6 × 2 pt (orange) configurations. Contours for the $6\times 2\mathrm{pt}:n_z^{\mathrm{shift}}:\mathrm{do}\mathrm{nothing}$ configuration are excluded from this figure owing to difficulties in reaching stable convergence within the inference chains, though the MAP estimate reveals a strong bias, at > 6σ on the S₈ − Ω_m plane (Table 6) The addition of clustering cross-correlations within the 6 × 2 pt configuration increases the precision of constraints upon the shift parameters, δz_i, thereby preserving > 90% of the idealized FoM whilst erasing the bias on the σ₈ − Ω_m plane (given in the legend). Conversely, the {γγ, 3 × 2 pt, 3^ph × 2 pt} : δz_i forecasts compensate for the parameter bias, but see reductions in the FoM by factors of ∼1.8 − 5.