This script spins up a number of agents (default 4), each with a certain average timeBetweenActions (default 400msec), and it uses a Poisson process in order to keep that average timeBetweenActions, but randomly distributed within that constraint. Think - popcorn popping!
This statistical trick is a valid model of random arrivals of users at a website, and will distribute load nicely. Thanks to the live-updating progress indicators of ascii-progress, and the awesome stream-fu provided by RxJS!
Here's a short video of my thoughts on the matter: https://www.youtube.com/watch?v=0GnB8FAJjfY