> 4) a new call is being triggered to data_window translator
Is your translator forkable (the default) such that it can be called asynchronously? What you may be seeing is that the Execution is not be closed (which can happen in a different thread) before the next Execution is being requested. You can confirm this by verifing that Execution.close is being called and that for subsequent reuse after the first you are getting an existing execution. This is a stricter lifecycle than what was enforced prior to the patch - which simply assumed the Execution was ready for reuse after reset and did not strictly enforce that the prior close would be called before reset.
If this is the behavior you are seeing, you could have your translator return isForkable false - which would make all interactions with the Executions happen synchronously in the processing thread. However if it's still desirable to use a separate thread for execution/results, then a more specific property, such as synchClose could be added.