Electrical resistivity tomography (ERT) is a technique widely used for the investigation of the structure and fluid-dynamics of the shallow subsurface, particularly for hydro-geophysical purposes, sometimes using cross-borehole configurations. The results of ERT inversion and their usefulness in solving hydrogeophysical problems, even though invariably limited by resolution issues, depend strongly on the accuracy of inversion, which in turns depends on a proper estimation and handling of data and model errors. Among model errors, one approximation often applied in cross-hole ERT is that of neglecting the effects of boreholes and the fluids therein. Such effects inevitably impact the current and potential patterns as measured by electrodes in the boreholes themselves. In presence of very saline fluids, in particular, this model approximation may prove inadequate and the tomographic inversion may yield images strongly contaminated by artefacts. In this paper we present a case study where highly saline water was used for hydraulic fracturing to improve permeability of a shallow formation impacted by hydrocarbon contamination, with the final aim of improving the effectiveness of in situ contaminant oxidation. The hydraulic fracturing was monitored via time-lapse cross-hole ERT. Arrival of the saline water in the monitoring borehole likely caused a strong borehole effect that significantly affected the quality and usefulness of ERT inversions. In this paper we analyse the experimental dataset and produce, via 3D ERT forward modelling, a viable explanation for the observed, paradoxical field results.