At a late stage in the construction of the plant for an on-shore oil field, a flaw was found in the circumferential weld of a thick walled header during routine non-destructive testing. TWI was called in to investigate. The matter was of considerable urgency because the header was part of the plant that separated gas from the oil stream. The gas was to be reinjected into the well. TWI's metallurgical engineers diagnosed the flaw as a fabrication weld metal hydrogen crack, but unusually it had extended into the parent forging by cleavage. This finding called into question the fracture toughness of all the forged components. Low toughness combined with flaws that might have gone undetected could imperil the plant. An extensive programme of Charpy V notch and CTOD tests indicated that the parent material toughness was indeed low at the design minimum temperature of -46◦C. It was originally supposed that such low temperatures might arise during rapid depressurisation. However, the operating oil company found that a more accurate and specific analysis indicated that the original design minimum temperature was excessively low. A more realistic minimum was -10◦C. A fitness for purpose analysis, based on BS 7910, showed that, at this temperature, surface flaws as large as 120mm long x 17mm deep were acceptable. It was considered that such flaws would be readily detectable and thus the plant could go into service without further remedial measures. This decision was extremely important to the operator, since it allowed the plant to come on stream without delay.