Anomalous fractionation of the minor isotopes of sulfur (Δ33S, Δ36S) in Archean pyrite is thought to reflect photochemical reactions in an anoxic atmosphere, with most samples falling along a reference array with Δ36S/Δ33S ≈ −1. Small deviations from this array record microbial sulfate reduction or changes in atmospheric source reactions. Here, we argue that reworking of atmospheric sulfur with distinct minor sulfur isotope ratios (Δ36S/Δ33S ≠ −1) produced additional variability in sulfide Δ33S and Δ36S-values in a 3.52 Ga hydrothermal barite deposit at Londozi, Barberton Greenstone Belt, Swaziland. In situ measurement of the four stable sulfur isotopes in pyrite revealed Δ36S–Δ33S relationships and a Δ36S/Δ33S trend (−3.2 ± 0.4), which is significantly different from the co-variation between Δ36S and Δ33S in the co-existing barite that reflects ambient Paleoarchean seawater sulfate. This argues against biological or thermochemical sulfate reduction at the time of barite deposition, and requires incorporation of sulfide generated in a chemically distinct atmosphere before 3.52 Ga. We propose a model that combines reworking of this sulfur by hydrothermal leaching, deep mixing with juvenile sulfur and surface mixing with biogenic sulfide to explain the observed variation in δ34S, Δ33S and Δ36S. These interactions between abiotic and biological processes in the Londozi hydrothermal system complicate the interpretation of biosignatures based on deviations in Δ33S and Δ36S from the Archean reference array.