I figured it out.
In my pipelines.yml I have multiple pipelines defined off the back of my CI platform.
In it was a test pipeline we abandoned that was using RabbitMQ. I left the config in there, but knew it to be abandoned. RabbitMQ provider was error'ing out. I thought nothing of it because, well - all the other pipelines run. One I@m not interested in is having an error, why would it affect another? Low and behold, disabling an erroring pipeline allowed the API to spin up on port 9600 as desired.
Odd behavior.