Torpedoes that missed did typically sink at the end of their run. However, they could be set to float instead. This was typically done with practice torpedoes. Use of this with real warheads was rare. The classic example of this is the actions of Lieutenant-Commander Martin Dunbar-Nasmith in the Sea of Marmora in 1915. Nasmith commanded HMS E-11, the second British submarine to make the dangerous passage through the Dardanelles. Coming to the end of his patrol, and wishing to maximise the effect of his last five torpedoes, he ordered his torpedo crews to set their weapons to float at the end of their runs. On the 28th May 1915, following a failed attack on a Turkish merchant, Nasmith would recover the first of these torpedoes himself. Diving over the side, he was able to disarm the torpedo's detonator, after which it was hauled aboard and maintained. This was a dangerous operation. Beyond the obvious risks from the torpedo, the operation required E-11 to remain on the surface throughout. The next (and last) recovery was carried out by Nasmith's First Lieutenant, Guy D'Oyly-Hughes. Disarming the torpedo's warhead as Nasmith as done, this time the warhead was pushed back into the sub's stern torpedo tube. This was slower, and harder for the men in the water, but let the sub remain in diving condition throughout. Torpedoes that sank when they missed might detonate on the bottom. A number of reports of submarine attacks mention explosions from torpedoes that missed. Those that did not explode may remain in dangerous condition on the bottom, but every work I've seen on unexploded ordnance at sea focus on the more common mines.