we should put in the docs somewhere our thoughts on memory limitations and when they'll arise. It will depend on the size of their data and the number of cores. I think we try to avoid high mem throughout the code pretty well. The biggest mem use is by vsearch in step6, which I believe is approximately the size of <name>_catcons.tmp * Ncores. So if that file is 1GB and you use 30 cores then it will use 30GB RAM. We could set an upper limit on Nthreads for vsearch to limit the max mem.