Is it v7.5.0 app server? If so, this is functioning as expected. USP was set to not share remote work items because of a massive usage of remote work items for that job type. This caused crashes of app servers due to out of memory for large scale jobs (10k+ targets).
Regardless of this change, performance of the job between 7.4.x and 7.5.0 is improved.
Post 7.5.0, we’re doing some optimization to the job that will make it much faster.
currently it's a 7.4.5 appserver
is there a guideline about sizing parallelism of USP jobs?