Nodes right now sometimes fork off a very small worker function to do some looping. In a pool of many nodes where individuals can be shut down quite often, this potentially leaves orphaned threads running. Each of those threads should have some place that waits for them, even if it is just a value in the environment that stores the cancellation actions.
Nodes right now sometimes fork off a very small worker function to do some looping. In a pool of many nodes where individuals can be shut down quite often, this potentially leaves orphaned threads running. Each of those threads should have some place that waits for them, even if it is just a value in the environment that stores the cancellation actions.