I have noticed the same behavior. This appears to be a Hangfire bug. It should not be trying to create a job for a queue that it is not registered for. Have you received any replies?
Neither received any replies nor could find the solution. For now, separate DBs are created to handle queues of each APIs (didn’t like this solution but had to implement).
You said that you have noticed same behaviour, so did you find any solution for the same?
I also wanted to use single DB for multiple instances running on different machines. As per the documentation here it is supported since version 1.5. You are using V1.5 is it still not working?