Last two minutes 

In spite of a common opinion, the pushing of SCRAM button is necessary not only in an emergency. On the contrary, usually SCRAM works in an emergency automatically, but the SCRAM button is often used as final operation at planned shut down of the reactor when its power is already enough lowered.
With reference to a situation on 4ChNPP on April, 26th 1986 there were the following. During all time after a downfall of power at 0:28 and before SCRAM button pressing at 1:23:40 there proceeded in the reactor and in the MFCC transitive thermohydraulic processes which can be qualified as unstable work. These processes are represented on the site in section "Actual data". So it seems very natural the operators wish to finish this work and to shut down the reactor as soon as possibly.
By the time of 1h 23min 40s the slowing down test has been almost finished (the TG slowing down lasted 35 seconds, a diesel generator has picked up 6 steps of loading from 9) and it was quite possible to stop the work. As SCRAM operation (in the test program) was assumed from the very beginning of slowing down there is nothing unusual in shut down the reactor by the SCRAM button at the end of test.
There is a question, why it have not been made half a minute earlier so as it was stipulated? After all, the reactor is unnecessary for the slow down of turbine-generator . But what is the sense of these 35 sec against 1.5 hours of hard work since the shift beginning, what such haste is for?!! After all the operation personnel did not know that they sit on a powder keg, and each second is decisive (if they will have luck).
There are also additional reasons why the reactor should be shut down. The last two minutes of transients in the MFCC (and in the reactor) are represented on the Pic. 20).

After the last great overcorrection of the feed water flow the flow rate strongly decreased, having returned on the initial value corresponding to power of 200 W. Simultaneously with it the pressure in separator drums (SD) fell till that of the moment of closing Isolating & Control Valves of the turbine. From the moment of closing ICV, the pressure has again increased, but it has begun the MCP slowing down and reduction of the flow rate through the reactor core.
All these processes (especially falling of coolant flow rate through the reactor core), and also permanent growth of the feed water temperature, should lead to occurrence in the reactor of positive reactivity. Speed of input of this positive reactivity could quite exceed the speed of xenon poisoning, and total reactivity could be positive, demanding to balance this reactivity by automatic control system. Well, such situation was represented in the official report for IAEA in August, 1986 by the Soviet experts, whose data of rods moving also is resulted on Pic. 20.
As the reactor power thus did not grow according to sensors of automatic control system, the positive reactivity was successfully balanced. And the [AR2] rods had not time to reach yet to bottom limit switch (BLS) by the time of SCRAM button pressing. But the there was already such situation that either it is necessary to turn to a hand control, or to shut down the reactor.

Anything, no objective data testifies to the version that operators have seen something meaning beginning of accident (and it has passed by automatics), and for this reason in haste have pressed SCRAM button.