This is really cool, thanks for sharing! I have recently been working on a workflow for using cmdstanr within the many-short-chains paradigm as explained in Margossian et al. 2024 on a cluster. Currently I use snakemake for this, but perhaps diststan would be a more accessible solution if it could somehow support the specification and bookkeeping of superchains (groups of chains with the same init values).
I think it does what you are asking. In one of the readme examples you’ll see you can specify more than one chain per machine (e.g., 2 for box one in the code below):