The Hyatt Regency Walkway Collapse

In July 1981 the tension columns supporting an elevated walkway inside the lobby of the Hyatt Regency hotel in Kansas City failed, allowing the walkway to fall four stories to the lobby floor. This remains the deadliest structural failure in American history other than acts of terrorism, and you can read more about it here:
The Hyatt Regency walkway collapse

It's usually hard to pin down a single reason why engineering disasters occur, but the causal chain is actually pretty clear-cut here. The following illustration included in the Wikipedia article is insightful and horrifying once you realize what it's showing:

The blueprints for the walkway design called for long columns to come down from the ceiling, through the the fourth-floor walkway beams, and terminate in the second-floor walkway beams. When contractors had a difficult time finding and working with columns long enough, the choice was made to split each column into two members each half as long, one reaching between the walkways, and one connecting the upper walkway to the ceiling. As the illustration shows, this doubles the force carried by the nut capping the lower side of the upper walkway-ceiling beam. Even in civil engineering, where large factors of safety are easy to apply to everything, doubling the load carried by a fastener without accounting for it in the design is completely and obviously unacceptable.

With the nuts' carriage loads doubled the fasteners were able to hold the dead load of the structure and occasional walkway traffic. The redesigned structure was completely inadequate when the walkways were flooded with spectators during a dance competition, though, and the nuts snapped off, allowing both walkways to fall to the ground at the least opportune time, when both were full of people. By the time the rescue was complete 14 hours later, 114 were dead and 216 injured because someone having a bad day forgot how free-body diagrams work.

How do events like this happen? That's not a rhetorical question. People train for years academically and professionally before our society allows them to sign off on the kind of life-supporting structural work that failed here, and this work is never (supposed to be) done without many people watching over each other, making sure that everyone's work is done in good faith with the physical world. The good news is that structural failures like this are very rare, and the industry does make some effort to learn from what went wrong. Hopefully this particular failure mode will never happen again.