Major Description
The current default are still too low for a large cluster and it's difficult to diagnose the issues when hitting these limits (there is also not really not much you can do but raise them). With the current thread model we use, we need to just raise these limits up very high.