This all sounds normal for batches of assembly jobs. These run on the largest busiest clusters. The jobs already running take longer to complete and to turn over for new jobs. That said, the resources are considerable, so getting those jobs queue and letting them complete is best strategy, and using collections and workflows will speed that queuing process up (along with the upstream QA steps, and your downstream post-processing).
How many people are working at the same server is also part of the calculation. This changes day to day and month to month.
More general details are at
Then, you can see the specific jobs and tools that are running at UseGalaxy.eu from a link on their homepage! → How to see the UseGalaxy.eu job queue statistics - #2 by jennaj. The other servers will look about the same. People find out where the throughput is lower and move over, then that one gets busy, the flow goes back, repeat. But I think just setting up shop on one server, getting the workflows ready, your data prepped, and the workflow invoked is still the best way to do this.
Hope this explains how things work, and that you have made progress since you asked! 