Just curious - assuming you pass the same tuning data to each chain, is many short chains equivalent to fewer long chains? Or are there other tangible benefits to running a large number of chains?
I’m wondering if it’s worth the effort to try and parallelize over a spark cluster.