job_conf_from_sorting_hat
Tools can require several CPU and/or some minimal of memory.
Since we have now more than 400 tools, it planned to be plainful to browse this list of tools.
So I decided to start from the sorting-hat project. Even if we didn't take the time so far to implement this project, we can use their list of tools and their "advices": https://github.com/usegalaxy-eu/sorting-hat/blob/master/tool_destinations.yaml
I translated using a dirty python script this .yml to a classic job_conf.xml (using a yml config + jinja2 template). I simplified the number of destinations by have some sorts of class with steps.
I also implement the resubmit by adding for each destination, a mem x 1.5.
It have to be tested on the preprodution. The dev doesn't have the necessary tools.