The quick drop from 725 MW to 512 MW in 22 seconds didn't happen. According to logs the power was gradually reduced by operators from 720 MW at 00:05 to around 500 MW at 00:28 in a span of 23 minutes.
RBMK reactor originally had automatic power control system which consisted of two automatic regulators AR-1 and AR-2 (one is active and another is backup) but it was not effective in controlling a reactor (an operator had to intervene every 10 seconds to control power level) so in 1983-84 reactors were updated with another automatic power control system called LAR (local automatic regulator) which used separate detectors and control rods but this system was effective only above 10% of nominal power (>320 MW).
According to RBMK documentation:
- AR managed 12 control rods
- LAR managed 12 control rods
- AZ (automatic emergency protection) managed 24 control rods
- PK-AZ (automatic protection overcompensation) 24 control rods
- PK-RR (manual control overcompensation) 12 control rods
- RR (manual control) 103 control rods
- USP (shortened absorber rods) 24 control rods
The reactor operating manual didn't restrict number of control rods pulled out but operators had to maintain so called ORM parameter (operating reactivity margin) which is calculated by the computer once in two hours or by a request from an operator and the results were delivered in 5-10 minutes. This ORM is represented as an equivalent control rod number and should be maintained between 30 and 15. If it gets below 10 then the reactor should be shutdown.
It is unclear if transition form LAR to AR happened automatically or was initiated by Toptunov. When AR-1 was activated it automatically moved twelve AR managed control rods to the upper limit stop switch due to imbalance in detectors signals and then deactivated. As the the AR managed control rods were already at the limit, AR-2 didn't activate. As all automatic power control systems were off, the emergency power increase rate protection system was activated and the operator reacted to it by pressing 'fast power reduction button' dropping the reactor power. (source: INSAG-7 report, section 1-4.7.3. page 73)
Dyatlov in his book says that this happened due to failure of AR. Tregub (chief of previous shift who was still present in the control room) says that Toptunov failed to handle transfer from LAR to AR properly.
Once the power dropped to 30 MW they (Toptunov and Tregub) immediately started pulling out RR (manual control rods) and probably pulled out almost everything to get power to rise again. Some say that it was Tregub who brought the reactor back and Toptunov just stood behind but Tregub at the trial insists that he was just giving advice to Toptunov. They managed to get it going in about 5 minutes and around 1:00 they stabilised the reactor at 200 MW power and engaged AR again. Around the same time ORM calculation was made with the result of 17-19 equivalent rods which is within the regulations.