If an evaluation is triggered during the job runtime (e.g. if a node is lost) we should refresh the list of allocations, as new ones may have been placed. This probably requires making a blocking query against the original evaluation, using modify index, in a separate thread.
If an evaluation is triggered during the job runtime (e.g. if a node is lost) we should refresh the list of allocations, as new ones may have been placed. This probably requires making a blocking query against the original evaluation, using modify index, in a separate thread.