Only certain protocols use those checkpointing flags. Most other protocols follow an output-level checkpointing approach. That is, they’ll make a list of what structures they need to output, and will skip running for structures that have already been output. So you should be able to re-start a stopped Rosetta job and it will pick up from the structure it last output. (By restart I mean rerun the exact commandline that you used to start it in the first place.) In principle this means that any work on a structure that hasn’t been output yet will have to be redone, but for most protocols this is 20-30 minutes at most, which is usually acceptable given normal running behavior.