It was during the morning rush hour when the departure signal at Waterloo station was giving an intermitent operation in that sometimes it turned green, but sometimes it didn't forcing the driver to proceed under the 'stop and proceed' rule applicable to underground working after a five minute wait. The signal engineer for the area was unavailable and so an engineer was summonsed from Kennington station. When he arrived, he gave instructions for the platform staff to hold the next train even if the signal turned green, while he entered the tunnel and looked for the fault.
He quickly found a signalling relay that although it was de-energised (known as 'down' in railway signalling speak), there was nevertheless a voltage across a normally closed contact (known as a 'back contact' in railway signalling speak), indicating that the contact was faulty. As a temporary measure to get the trains moving, he shorted out the contact with a piece of wire terminated in two crocodile clips. The engineer was mindful of the fact that he had substantially shortened the overlap distance of the signal. When he arrived back at the station he instructed the platform staff to hold every train for one minute after the signal turned green. He then went in search of a second wire with the intention of temporarily connecting the faulty contact to an unused contact on the same relay restoring normal operation. It was while he was searching that the collision took place just outside Charing Cross station.
In the enquiry, the signal engineer was held wholly to blame, because although his temporary fix was reasonable under the circumstances, the instructions that he gave to the platform staff were incorrect. The holding of the train for one minute after the green aspect was reasonable, but the driver should have been instructed to, "... proceed at caution and be prepared to stop short of any obstruction". The absence of the last instruction meant that the driver would proceed at full speed and be unable to stop short of another train.
The inquiry and final report had an unexpected twist, because the engineer offered in his defence that he had introduced the equivalent of a single point failure, exactly what would have happened had the contact stuck closed. He pointed out that it was a legal requirement that signalling systems should be immune from single point failure modes, and the board of enquiry had no alternative but to agree.
As a result, the berth circuit (not to be confused with the 'berth track circuit') was developed whose purpose was to detect contacts that were closed when they shouldn't have been and to rupture the relay supply fuse if such a condition was detected forcing the controlled signals to show an unchanging danger indication. The berth circuit was retrofitted to every automatic signalling system in Britain.