You are using explicit dynamics step with a specified minimum time increment which means that mass scaling is used. It can be large and cause jumps in kinetic energy, affecting the results.
Your analysis submitted in Abaqus with mass scaling using the same target time increment as your CalculiX simulation (2e-5):
The same analysis but with two orders of magnitude smaller target time increment (and thus much lower mass scaling):
Plots on the right show internal and kinetic energy.
I would recommend lowering the minimum time increment or (even better) switching to dynamic implicit or static analysis and using prescribed rotational displacement instead of initial velocity. If it doesn’t converge, you can go back to explicit but keep in mind that you are applying initial velocity which means that the gear is spinning initially, only to hit the other gear (in your main simulation) and stop rotating eventually. It’s a highly dynamic even, basically impact problem. So you should decide whether you want to simulate something like this or constant rotation during operation of the gear. In the latter case, you will need a boundary, not initial condition and explicit analysis may not be necessary, as explained above.
P.S. I made a tutorial about explicit dynamics in CalculiX as well, covering the importance of the incrementation settings: https://www.youtube.com/watch?v=-X0Shj7UXE0