I'm thinking of taking an existing Snakemake-based system running on a local SLURM-supported grid and moving it to AWS. A single run may span ~12 hours involving dozens of cores. Any ideas on if getting a stable solution would be faster by switching to Grail: reflow? Or should I gut it out and add in cfncluster myself? Any one with experience with either direction, i'd love to hear your experiential wisdom :) Thanks