My company found Batch API (JSR 352) useful but the design with Beans in JSR 352 is horrible.
Therefore I am proposing to deprecate it and to create a new one.
There might be voices, as mine, saying that the Steps of the Batch should be Cloud capable and the Steps should be able to fork their execution in Docker via Cloud orchestrator.
Anyway this should be reworked and get the attraction in the future.
So altogether two approaches may exist 1. Steps in one JVM and 2. Steps distributed over multiple JVMs.
Regarding the JSR 236 is very important but here is fundamental problem of the designers because they made the Beans a non-beans so that they are totally cutt off the EE context. Therefore I found this API nice but incomplete as it is a good theoretical API with no practicle application in EE Beans managed world.
Anyway the managed threal pool is important from the configuration perspective and therefore manageable. But we require more than this and so the Thread should fully understand the EE context.