It was exactly 19 years ago. Soviet Union was trying, at least in theory, to reach the perfect communist society. Since May 1 (Labor day in USSR) was close, the staff at Chernobyl nuclear plant was rushing to complete one more final test that would serve as a proof that their power plant would survive a sudden energy blackout. The goal of the test was to show that the nuclear reactor #4 can be "rebooted" in the absence of any external electrical energy sources.
Nobody knew that the test (as designed mainly by Nikolai Fomin, the plant chief engineer back then) was pretty flawed as it required the reactor to work outside its designed operational limits. What the designers did not know was that for the duration of the four hour test, the reactor was also in a highly unstable state, due both to its positive void voeficient (i.e. nuclear reactivity goes up when steam is formed in the cooling water) and positive temperature coeficient (i.e. reactivity increases with temperature). Sadly enough, this information was actually classified back then, in the true spirit of the communist society. Not even the people operating the plant had access to this information.
There were also a number of violations in operating procedure. First of all, the test required the automatic control system (LAR) to be switched off. Second, at the initiative of the chief plant engineer, the emergency core cooling system was also turned off for the duration of the test, based on the fear that it will be dangerous if the cooling water from the emergency tanks enters into the hot reactor. Finally, both emergency diesel generators were turned off, in order to simulate a "pure" experiment. Even with these stunning violations, people were confident, since such tests were conducted previously with success, so nothing could wrong this time, right? The reactor was generally viewed as safe as a rock.
During the test, however, several things went wrong. First, with LAR switched off, the operators were unable to manually maintain the reactor power under the devised limits. The normal (operating) power was around 3200 MW thermal, which meant about 1000 MW of electrical energy produced. The power required during the test was much lower - a few hundred MW, simulating the reactor startup conditions. However, the reactor went to 10x lower than expected - 30 MW thermal.
At this point, theoretically, the reactor should have been shutdown, and a regular start-up been scheduled the next months. But an immediate restart was attempted as a quick "cover-up" for the earlier mistake. But that didn't really work. The start had been somewhat impeded by a known fenomenon called Xenon poisoning. (Xe-135 is a fission decay product that is a strong neutron absorber. In normal operating conditions, the produced Xe-135 is held in check by the neutrons produced in excess. But during shutdown, the concentration of Xe-135 dramatically increases making the reactor restart impossible. But since its half-life is about several hours, the reactor restart will be again possible after a few hours. More interesting details here).
The test went on, under the impatient management of the deputy chief engineer - Anatoly Dyaltov. He ordered that, one by one, the remaining control rods to be retracted from the core, to move the reactor back into the operating position. Anyway, the rods were withdrawn until the reactor was raised at approx. 200 MW thermal, with very few remaining control rods in the core. But he didn't know that at this point the reactor was in extremely unstable state. After few minutes, everything looked stable and, around 1:23:00 AM, he decided to start the actual test. As we briefly mentioned above, the test required that the massive turbogenerators would be started on using only the hot steam produced from the core. But when the turbogenerators were connected in the flux, however, the flux of incoming water started to decrease. Bubbles of steam started to form into the core coolant, which caused sudden increase of power due to the positive void coeficient.
The operators noticed this, and at 1:23:40 AM they tried to manually shutdown the reactor. However, when the safety rods were introduced in the core, the reactivity actually increased for a few seconds due to a fault in the design of the rod geometry. This was the final blow that brought the reactor in a few seconds to around 30,000 MW (thermal). The huge generated heat caused several large explosions that blew away the 1000 ton reactor lid and destroyed the top of the building. Due to the lack of containment, the core was now directly exposed to atmosphere, releasing hundreds of millions of Curie in the environment for the next days.
The reactor was "shutdown" in the next weeks, by dumping sand and lead on it. At the end, only around 10% of radioactive material was left in the building, the rest was released in the meantime by the explosion or the intense fire.
31 people died, mostly firefighters that heroically fighted with the fire on the roof of the building. Most of the operators died too. Hundreds of people were hospitalized, many of them dealing with acute radiation problems. But the long term consequences were much more severe. Hundreds of thousands of people were relocated from the the affected area, with minimal support from the government. They had to start their life from scratch, with no resources, no past.
[update - corrections]