Sorry, the answer was a bit terse, knife exec just run ruby, with all chef classes available.
What it does here is searching for nodes, and on each node object it got, modify it's run list attribute, roughly what can happen in a chef run on a node.
When a node run is successful, the node object is saved back to the server, that's the call to node.save here.
That's a bunch of inner API of chef, but the doc is here https://www.rubydoc.info/github/opscode/chef/Chef (more or less complete depending on the class).
For the parallelism part, I don't think it would be short enough to declare as a one liner, you'd need to code it in pure Ruby, handling threads, etc.
knife exec -E 'code' is roughly running a ruby script with chef and knife classes already loaded.