Hello,
It could be also that you have von Karman eddies behind the island, in which case such variations of velocities are normal, and the local turbulent diffusion is a sensitive factor, this is another hint. Normally we would expect that such eddies are triggered by the island, this can be observed in reality (this does not explain the difference between two versions if you do not use new features, however these oscillations may take time to settle and depend on bifurcations, so it can be sensitive to slight changes).
Regards,
JMH