The multiscales parallelism didn't take into account downsampling when creating the spawned solvers, so those solvers had the wrong scalelo and scalehi, and then they might reject needed index files. This issue doesn't happen with other forms of parallelism, just with multiscale.
The multiscales parallelism didn't take into account downsampling when creating the spawned solvers, so those solvers had the wrong scalelo and scalehi, and then they might reject needed index files. This issue doesn't happen with other forms of parallelism, just with multiscale.