The parsl+wq and parsl+taskvine test suite takes 493s to run, vs 8s for 
the same tests run in parsl+htex.

Right now that's the dominant time in our CI - lets say 15 minutes out of 
a 22 minute CI runtime.

I think that's coming from sequential task handling somewhere in the 
WQExecutor/WQ stack taking around 1 second: where task B is not submitted 
until the parsl DFK has become aware of task A completing. (task A being 
one test, task B being the next test, or in some cases task B being a task 
inside a test that depends on task A completing).

Is there a way to make this faster? In principle, parsl+wq coprocesses 
should be down at the 8s mark too.

The bit of CI that does this test is this:

pytest parsl/tests/ -k "not cleannet and not issue363" --config parsl/tests/configs/workqueue_ex.py

(that's the same test suite parameterised by different test 
configurations, so --config parsl/tests/configs/htex_local.py will get you 
the same tests but through the high throughput executor)