Training currently blocks the counting of new laps using new detections, If this would happen in the background, the potential delay would be smaller