Exam DP-750 Topic 1 Question 69 Discussion

To achieve this, you should set a "Failure" notification at the individual task level for critical tasks only, rather than relying on a job-level notification.
Task-Level Notifications (Critical Tasks): By configuring notifications exclusively on the critical tasks for the "Failure" event status, any failure of these specific tasks will immediately send an alert. Because tasks execute within the job lifecycle, this notification inherently triggers while the job is in progress or completing its critical paths.
No Job-Level Notifications: You must not set a job-level "Failure" notification. If a non-critical task fails, it can cause the overall job status to register as a failure or "Succeeded with failures", which would erroneously trigger a job-level alert.
Handling Non-Critical Tasks: Non-critical tasks should be left without task-level failure notifications. To ensure they don't break downstream flows, you can configure downstream dependent tasks to use the "Run if: All done" or "Run if: None failed" (depending on your dependency architecture) conditional rules so the job execution continues and logs the failures automatically to the matrix view.
Reference:
https://learn.microsoft.com/en-us/azure/databricks/jobs/notifications