TKE strange values while running in parallel

While working on my test case with drag force and offshore structures (that was finally working), I wanted to make a reference simulation without any structures (no drag, no TKE source term => iopt_drag_3D=0) and a strange behavior happened in the turbulent kinetic energy (tke) output variable.

• Some points of 'high values' of tke appear at the corner of the sub-domains (40 processors) : see screenshots below.
• I tested with the last main branch and it appears directly (first timestep) and on the upper sigma layers (first 3 or 4 layers from the surface). Weird...
• Variable Kz (vdifcoefscal) seems okay.
• 
•