Request for a top-level handler, or a reserved branch on workflow definitions.

Today, there's no way to reliably trigger a notification (Slack, PagerDuty, etc.) when a workflow fails without compromising the workflow's final status. The current options available are:

onFailure: terminate — correctly marks the workflow FAILED, but downstream notification nodes never run.

onFailure: continue — notification nodes run, but the overall workflow is marked SUCCESS despite the upstream failure.

Impact: Can't easily implement basic alerting without misrepresenting operational health.