Very well written.

I think using 'make' with the -j parameter (# of parallel jobs) is more useful than using gnu 'parallel'. The reason is that if one of the job fails for some reason, you just re-run 'make' and only the required jobs are started instead of restarting the entire computation.