We should build out our ability to stop and restart (or restart from failure): * write the convergence file after each major cycle * implement an option to copy prev -> current and maybe validate that prev's clean flux matches the convergence file
We should build out our ability to stop and restart (or restart from failure):