“…On the other hand, building a coarse solver with explicit time integration cannot be based on an increase of the time step, because of their intrinsic CFL limitation. One can use then coarsening in space, but previous analyses in the lit-erature [29,33] show that one needs necessarily high order interpolation between the coarse and fine grids, and the coarse solver could then also suffer major efficiency loss in the context of massive space parallelization, not even talking about further high frequency error components introduced due to this process, which the iteration can not eliminate. This makes it very tricky for actual PinT methods based on multilevel techniques, and it seems unavoidable that one has to develop new techniques for the coarser levels when applying Parareal like methods to the advection problem, or more generally, hyperbolic problems.…”