These are chat archives for dereneaton/ipyrad
I can see how branching and then merging data sets could be useful. For example, if you had a mix of haploid and diploid samples that you wanted to treat differently. We can work on making this possible, but we've discussed it before and it's a bit complicated if we allow data sets to be merged at any step. It is currently easy to apply step functions to a subset of samples in an Assembly when using the API, but that is not well documented yet.
As for speed, I don't think branching and merging should typically help for speed purposes. On an HPC cluster you should be able to connect to as many cpus as you can get and the speed improvements across samples should be nearly linear. Just make sure you use the --MPI flag when connecting to multiple nodes on HPC.