Chaos Monkey
Netflix tool that randomly kills production instances to force the team to build systems that tolerate node failures.
Launched 2011. The philosophy: if your servers die during work hours with the team on hand, that's much better than at 3 AM on a Sunday. Part of the "Simian Army" — Latency Monkey (injects delay), Conformity Monkey (kills non-conformant instances), Chaos Gorilla (takes out a whole AZ).
Birthed the field of chaos engineering. Modern tools: Gremlin (proprietary), Chaos Mesh (CNCF), Litmus, AWS Fault Injection Service. Toxiproxy is its network-layer cousin.