These tasks tend to be very short but can queue behind grade reports

Here are the 3 tasks over the last 30 days with average duration
runtimes in seconds:

lms.djangoapps.instructor_task.tasks.calculate_students_features_csv 37.794393
lms.djangoapps.instructor_task.tasks.calculate_may_enroll_csv   0.030303
lms.djangoapps.instructor_task.tasks.export_ora2_data   67.629921

All but ora2 seem fine to pass through the default queue, which has 3
workers per physical machine.

lms/djangoapps/instructor_task/tasks.py

@@ -193,7 +193,7 @@ def calculate_problem_grade_report(entry_id, xmodule_instance_args):
     return run_main_task(entry_id, task_fn, action_name)
 
 
-@task(base=BaseInstructorTask, routing_key=settings.GRADES_DOWNLOAD_ROUTING_KEY)  # pylint: disable=not-callable
+@task(base=BaseInstructorTask)  # pylint: disable=not-callable
 def calculate_students_features_csv(entry_id, xmodule_instance_args):
     """
     Compute student profile information for a course and upload the
@@ -252,7 +252,7 @@ def proctored_exam_results_csv(entry_id, xmodule_instance_args):
     return run_main_task(entry_id, task_fn, action_name)
 
 
-@task(base=BaseInstructorTask, routing_key=settings.GRADES_DOWNLOAD_ROUTING_KEY)  # pylint: disable=not-callable
+@task(base=BaseInstructorTask)  # pylint: disable=not-callable
 def calculate_may_enroll_csv(entry_id, xmodule_instance_args):
     """
     Compute information about invited students who have not enrolled
@@ -293,7 +293,7 @@ def cohort_students(entry_id, xmodule_instance_args):
     return run_main_task(entry_id, task_fn, action_name)
 
 
-@task(base=BaseInstructorTask, routing_key=settings.GRADES_DOWNLOAD_ROUTING_KEY)  # pylint: disable=not-callable
+@task(base=BaseInstructorTask)  # pylint: disable=not-callable
 def export_ora2_data(entry_id, xmodule_instance_args):
     """
     Generate a CSV of ora2 responses and push it to S3.