Yes, some errors would instead appear right away when starting the simulation. This is why I think it may be related to the calculated water level.
If you're using the rectangular grid version of MIKE 21 (not the flexible mesh one), if your boundary cells are initially dry, then the error will be detected only when water will reach the boundary during the simulation. The error could be that, when this occurs, only some of the boundary cells become wet, which is not allowed. To avoid this, you may apply a constant bathymetry along the boundary, and also change the initial condition so that the cells along the boundary are all initially wet.
Best regards,
Mathieu
