A pollution peak was recorded in the discharge into the Seine river from the treatment plant of a refinery. This peak was the consequence of several malfunctions not detected or anticipated in time and partly caused by rainfall events. A level problem on a furfural-water azeotropic mixture separation tank on the FURFURAL unit caused furfural to be sent to the bottom of the water stripping tower. Colorimetric analyses were carried out on the aqueous effluents at the outlet of the bottom of the tower every 2 hours. The analyses carried out at 3 am showed no anomalies in the amount of effluents at the outlet, but a 2nd analysis carried out at 5 am revealed a peak of COD > 1,000 ppm. After 6 am, the bottom of the stripper diverted towards the refinery’s slop tanks in accordance with the existing procedure. As the COD peak at the tower’s outlet was only detected at 5 am, furfural was carried directly to the oily network of the refinery’s treatment plant for 2 hours. Heavy rainfall the night before between 5.30 pm and 9.30 pm, caused a significant amount of water to be sent to the plant’s gravity separators, along with the filling of the storage basin downstream of the plant. It was then impossible for the operator to divert the effluents coming out of the separators towards the downstream storage basin. As the storm at 7 pm caused the TOD meter upstream of the gravity separators to become saturated, the device switched to scrubbing cycle and transmitted an inaccurate value to the control room. The problem on the analyser was detected at 7 am during the technicians’ round. As the TOD meter measurement had been inaccurate most of the night, the COD peak at the API separator inlet could not be detected and the aqueous effluents polluted with furfural could not be diverted during the night towards the buffer tanks. After 9 am, the problem on the TOD meter was corrected and the effluents at the API separator inlet were diverted to the storm tank; a proportion of the effluents at the flocculation-floatation treatment outlet was able to be diverted to the downstream storage basin (100 m3/h). The aqueous effluents polluted with furfural were only diverted to the buffer tanks in the morning, and a proportion of the pollution was discharged into the Seine river (5.7 t of furfural). The inspector of classified facilities went in the field on 27/10/2006. The operator planned an intervention on the D4A tank in operation in order to regulate the furfural-water interface (18/10/2006), shutdown the furfural extraction unit to clean the D4A tank, unclog the dip tube (25/10/2006) and, at the same time, take hourly measurements on the effluents coming out of the bottom of the tower (unit restart on 26/10/2006), as well as revise the faulty measurement alarms on the TOD meter at the separator inlet.