We always hear about how they were running the reactor in a dangerous mode to run the experiment, but the nature of that experiment is rarely mentioned. The experiment itself is the craziest part!

The experiment was a test of a new cooling system feature. The RBMK's suicidally stupid design had a positive void coefficient of reactivity[1]. Coolant voids increased reactor activity. The operators new this; they had backup diesel pumps that would take over from the electric pumps if power wasn't available. They also knew that it would take some time[2] for the backup pumps to reach full speed, which was not fast enough.

So someone had the "clever" idea to modify the electric pumps so they would continue to spin freely when they lost power, so their remaining kinetic energy could continue pumping coolant at a rapidly falling rate. The hope was that there would be sufficient kinetic energy to keep the reactors from voiding. The experiment was basically a bad hack to work around several serious design problems.

The experiment shouldn't have existed. Everything about the coolant situation should have been a reason to decommission the reactor. The experiment itself is evidence that someone knew about the positive void coefficient problem, or they wouldn't have tried such a crazy workaround. Yes, a lot of the staff was poorly trained, but someone knowledgeable about reactors decided they couldn't afford even a few seconds[2] without coolant.

[1] https://en.wikipedia.org/wiki/Void_coefficient

[2] like 50 seconds? if I remember correctly?