Since the run is so short, the miliseconds really matter. The run is recorded at 60 frames pre seconds, and I timed it at 0,125x the speed. Turns out I started 70ms too late and split 10ms too late.

I also did this for the current world record, which started perfectly but finished slightly to early, too see how much world needed a timer to run with it.

Proof: Here it's visible per frame: