Can a turbulence experiment be repeated ?

Using time-reloved PIV, we make movies of a 2D slice of the velocity field behind an impulsively started submerged plate in a water basin. The experiment is exactly repeated 42 times using a robot. The flow is characterized by vortices. The two counter-rotating vortices that are shed initially merge and become turbulent. The outcome of repeated experiments diverges as time progresses. This divergence can be predicted from a single experiment by computing its finite-time Lyapunov field. Remarkably, the initial symmetry of the opposing large-scale circulation remains conserved over the entire trajectory of the plate. Yes, a turbulence experiment can be repeated, but not its outcome. However, in this experiment, the discrepancy between results takes on a remarkable shape.