If this is set, along with max_clusters_in_pool we should wait a random amount of time between 0 and this many seconds and double-check that no other clusters were created before launching a cluster of our own. This way, if a bunch of jobs launch simultaneously, we can make it unlikely that they will all create a cluster at the same time.
This applies to bailing out early from pool_wait_minutes as well; if we think there are no clusters to wait for, we should double-check before launching our own.
If this is set, along with
max_clusters_in_pool
we should wait a random amount of time between 0 and this many seconds and double-check that no other clusters were created before launching a cluster of our own. This way, if a bunch of jobs launch simultaneously, we can make it unlikely that they will all create a cluster at the same time.