For a Tracking Event Processor (TEP) to be able to do any work, it requires the claim on a Tracking Token.
Thus, a claim on the tokens stored in the token_entry table.
Through this claim, the TEP can coordinate work between distinct instances of the TEP or different threads present on the same TEP.
So what you are noticing, is the fact you only have a single tracking token, thus only segmented the event stream into a single set.
This means that two distinct nodes, both running the same TEP with a single thread, will still run single threaded.
If you want to multi-thread this process, you will have to segment the work into distinct portions by segmenting the tracking token into several pieces.
I’d recommend having a look at the reference guide on the matter, and start reading from here, to get a feel how to achieve this parallelization.