It would be nice if we could automatically scale down jobs which haven't been requested for X days. Trying to run 600 unused containers is not fun.