Client side may fail to get job status after invoking execute_query()

**Describe the bug**

Due to #134, at the scheduler side, we can not ensure that the execution graph created or there's a job info persisted to the backend after the execute_query returned. Then there's a chance that the client fails to get job status.

**To Reproduce**

**Expected behavior**

**Additional context**

Start by tracing the client-side execute_query() flow alongside the scheduler path described in issue #134, focusing on when the execution graph and job information are persisted to the backend. Reproduce the timing race and verify that the client can reliably retrieve job status after execute_query() returns.

Written by the indexing model from the issue text.