In the early days of nuclear power there was a misunderstanding of what caused power levels to change in a reactor...[pause for technical explanations]
Criticality is the measure of how the reaction rate is changing. In a critical reactor, on average, one neutron produced from a fuel atom fission will go on to induce another fuel atom to fission, creating a sustained chain reaction. If more than one, on average, is doing this than the reactor is supercritical and power levels are rising. The departure from criticality is measured in reactivity units, so a critical reactor is at zero reactivity, and you can make power levels increase by doing something to either add positive reactivity or remove negative reactivity, with the generally most direct access being withdrawal of control rods to remove their negative reactivity.
[end pause]
...anyway, what was misunderstood was that power levels not only respond to reactivity, they respond to the rate of change in reactivity. Since control rods are generally moved by motors that run at a fixed speed no one really noticed that part of the power level change was because of the increased reactivity that was added and part was from the actual process of adding the reactivity. For all (well, almost all) practical purposes it worked out the same anyway.
So, there's this tiny little test reactor, and for maintenance purposes the motors that drive the control rods have been removed. Now they are all spruced up and ready to be put back on. The control rods are fully inserted, so the reactor has a bunch of negative reactivity and is at some minute steady power level where stray environmental neutrons are enough to make up for the fact that far less than one neutron per fission is surviving to cause another. And a guy is tasked with hooking up the drive motors.
He goes to connect one, and the gear teeth don't line up with the ridges on the stem of the control rod. Well, actually they DO line up with the ridges, so the motor can't drop into place. The teeth need to be in the grooves between the ridges. So our guy figures "I'll just grab the stem and give it a little yank so the teeth can mesh."
Now, this reactor is subcritical by a huge margin, with a whole lot of negative reactivity, so moving this control rod and taking out a very little bit of that negative reactivity should not be a problem. However, the yank of the control rod removed that small amount of reactivity very very quickly, creating a very high reactivity addition rate and taking the reactor supercritical. It responded by producing a very high power level in just a handful of neutron life cycles, boiling the coolant to rapidly expanding steam that began expelling the control rod so rapidly that the teeth couldn't engage, creating an even faster reactivity addition rate, etc, etc, etc.
In the end this little test reactor, designed to produce several watts of power, produced some hundreds of kilowatts, perhaps even as much as a megawatt...for a very, very short period of time.
The ejected control rod ended up protruding through the several foot thick concrete roof of the containment building, leading to some consideration of the "nuclear gun."