so in this case that might mean "run locally for 5...
# development
w
so in this case that might mean "run locally for 500ms, then launch a remote copy to race against"
a
But in our context, if we always try to schedule something locally first, it’s always going to complete locally first, given the extra cost of remote is start-up overhead?
w
not if there is actually extra concurrency available
you attempt to schedule 100 things... four of them get local slots, the next 96 wait more than X ms, and get scheduled remotely
a
aha, queue was part of the run there
w
yea