An alternative to keeping the queue in a database would be to keep it in a plain text file, and manage access to the file with standard file locking flock etc.

In general though, I think that if you have a suitable database available and you are familiar with it, then that would be the more scalable and reliable solution.

I don't think you should create a separate table for each parent PID though as that will pollute the database with lots of tables that have the same purpose. If I where you I would share the same table between all workers, and just have a row that records the PID of the parent. Then in each worker the select call can filter on that PID and only get jobs that pertain to it.