The advantage of QSl is its atomar worker approach. We can spin up a defined number of workers and be sure that no internal mechnisms do things without our control. However, to be flexible with load balancing we need e mechnism to scale the amount of workers dynamically on metrics.

For instance, if the job queue lenght is:

• growing in average over X seconds we want to scale up to Y workers
• stable over X seconds we want do scale to a default number of workers

This is possible because each worker has a nearly instant boot process and can start picking jobs right away.