workflows

Automation turned off after errors: diagnose before restarting

Check the workflow state, first failure and pending effects in Make, Zapier or Power Automate before restarting or replaying work.

When an automation stops after errors, switching it back on can revive the cause or repeat effects that already happened. First distinguish a disabled workflow, a failed run and a pending item. Each needs different evidence.

Short answer

Confirm the current state and who or what changed it. Find the first failed run and its original error, rather than relying on the latest alert. Correct the cause, reconcile each affected item with its destination, and then restart with a controlled input. Do not replay everything when a message, record or payment might already have been created.

Restart decision matrix

What you see Minimum evidence Before restarting Before replaying items
Manually disabled workflow Change history and owner Confirm the reason and authority to restart Check work handled by another route
Workflow suspended after failures First failing action, time and suspension rule Fix the cause and test one controlled input Separate no-effect failures from uncertain effects
Broken connection Account, permission and failing action Restore the authorized connection Check the destination by business ID
Provider outage Vendor status and last confirmed result Wait for recovery and limit new inputs Avoid replaying calls with unknown outcomes
Incomplete run or queue Item ID, completed step and current state Assign an owner to resolve the queue Continue only from verifiable evidence

This is Pulse’s editorial decision model. State names and replay options vary by platform and workflow configuration.

Where to look in each platform

In Make, unhandled validation errors can end a scenario, and consecutive errors can disable scheduling after the configured threshold. Inspect the run history, error settings and incomplete executions before enabling the schedule. An error does not necessarily create a recoverable item.

In Zapier, “Task was stopped because its Zap was turned off” can mean someone switched the Zap off during a run, too many errors occurred, or an app or Zapier had an outage. Zapier’s guidance is to inspect and fix the errors, turn the Zap on, and consider replay afterward. The message alone does not establish the cause.

In Power Automate, “Off” and “Suspended” are distinct: Microsoft’s documentation distinguishes manual deactivation from automatic suspension after repeated failures. Check the flow state and failing run. A data loss prevention policy issue may require an administrator; do not bypass the policy to force a restart.

Recover item by item

For each business ID in the affected interval, record only what is needed in an authorized workspace: source, run, last confirmed step, destination effect, decision and owner. Use four states:

  1. No effect confirmed: fix the cause and consider a controlled run.
  2. Effect confirmed: do not repeat the action; reconcile the record.
  3. Effect uncertain: inspect the destination or provider before replay.
  4. Item changed later: review the current version before overwriting it.

Hypothetical example: three orders arrived before suspension. Order A never started; B created a CRM record but failed while sending a notification; C timed out without a receipt. After fixing the cause, A can move through a controlled path, B needs a decision only about the notification, and C stays under investigation. Replaying all three as untouched work risks duplicates. Pulse did not run this scenario as a test.

Microsoft warns that resubmitting entire runs can duplicate effects. Also confirm that the input remains valid and that the flow has not changed since the original run.

Stop condition and next steps

Do not restart while the cause persists, the affected items cannot be identified, an external action might have succeeded without a receipt, or you lack authority to handle the data or credentials. Use a manual fallback with an owner and reconciliation if the business outcome cannot wait. If the workflow is on but records are missing, trace source, run and destination. For an authentication failure, start with the 401 connection diagnosis.

Sources