On September 30, 1993, at about four in the morning, an earthquake measuring 6.2 struck the districts of Latur and Osmanabad in the state of Maharashtra, on the Deccan plateau of western India. It lasted under a minute. It killed about 10,000 people, injured 30,000, and left some 100,000 without homes. The village of Killari was destroyed, and with it most of the people in it who were indoors at the time.
The timing was the reason. It was early morning, the monsoon was ending, and people were asleep inside houses made of rubble stone set in mud mortar with roofs of heavy stone slabs laid across timber beams. Those roofs weighed tonnes. When the walls gave way the slabs came down onto the beds beneath them, and there was no warning, no foreshock, and no way out for anyone who was not already standing in a doorway when it began.
The location was the puzzle. Latur sits in the middle of the Indian plate, hundreds of kilometres from the Himalayan boundary where India's earthquakes were expected, and the official seismic zoning map put the whole Deccan region in the lowest risk category. This was an intraplate earthquake, on an ancient fault hidden under a cover of basalt lava that geologists had assumed was stable, and forty seconds proved otherwise.
Rescue came slowly. Army units arrived, and international teams came with search equipment and dogs, but the villages were scattered, the roads were poor, and the monsoon had softened the ground. Most survivors were pulled out by their neighbours in the first hours, with bare hands. The government paid compensation, set up a reconstruction authority, and began building new settlements on ground away from the worst-affected sites, with reinforced masonry to a standard the old houses had never met.
What changed afterwards was the map. India revised its seismic zoning in 2002 and moved the Deccan plateau into a higher category, and the Maharashtra Emergency Earthquake Rehabilitation Programme, funded with World Bank money, rebuilt hundreds of thousands of houses with the reinforcement that had been missing. The earthquake remains the standard case for why intraplate events are dangerous: they are rare, they fall outside the mapped zones, and the buildings above them were never built for one.