Choose one process and a harmless sample case. Agree these five decisions before testing your response to a failure. The worksheet stores observations only; it does not connect to your accounts, watch workflows, stop queued work or perform a retry.
Keep one small recovery record.
Use the PDF for a quick decision and the editable file for the evidence timeline. Record private references rather than names, passwords, message bodies or access keys. Keep the same case reference when responsibility changes.
Fictional heating-company example, not a tested customer integration
Case
Confirmed
Missing or unknown
Next decision
HEAT-08
One enquiry and one received confirmation email
Staff call-back task missing
Hold further attempts; owner verifies whether one task can be completed separately
A successful log entry and a visible result answer different questions. Where possible, check the actual conversation, booking, task or connected service before authorizing recovery.
Use current logs, then make a human decision.
In HighLevel, open the relevant workflow and use Execution Logs → View Details for the affected action. The official log guide explains the available history. The planner fields here are manual writing prompts, not configured software settings.
HighLevel documents a Review workflows with errors view in Workflow Maps. Mark as read acknowledges an error indicator; it does not fix the workflow or retry an execution. Check alert coverage and recipients before relying on email as your only watch system.
Hand over a decision, not just an error screenshot.
State the last confirmed outcome, the missing or uncertain action, verified containment, one approver and the next review time. If a result remains unknown, hold and escalate. After a fix, check both a controlled sample and the original affected case. A successful new sample does not complete an older missed task automatically.
Use a controlled sample and compare its private reference, workflow, affected action, time window and timezone with the evidence. Add a second similar sample so the reviewer must distinguish them.
Expected outcome
The owner identifies the correct case and action. A similarly named record or a different time window is not used as evidence.
TEST 02
Stop requested is not stop verified
In a harmless test or tabletop exercise, ask the operator to hold only the agreed work. Record the request and the evidence of the actual held state separately; list anything still unverified.
Expected outcome
The actual containment state is checked. Unknown queued or linked work is escalated; an edited worksheet is not treated as a stop.
TEST 03
Completed work stays completed
Model an enquiry where the confirmation email arrived but a staff task is missing. Compare both destinations and ask the approver to identify only the missing action.
Expected outcome
The received email is preserved as completed. The missing task has a separate recovery decision; the whole sequence is not replayed.
TEST 04
An uncertain result stays on hold
Use a tabletop scenario in which a connected service times out and its records cannot be checked. Ask the owner what they can honestly conclude and who can obtain the missing evidence.
Expected outcome
The outcome remains Unknown. No blind retry is approved while the destination cannot be verified; one owner seeks the evidence.
TEST 05
One approved recovery attempt
With harmless sample records, have two staff receive the same request. Ask them to find the shared case and verify who alone is authorized to perform the exact recovery.
Expected outcome
One operator owns the approved action. Others check the same case first; a repeated request does not become another attempt.
TEST 06
Backup and observed closure
Make the main owner unavailable in a tabletop exercise. Have the backup accept the case, state a review time and identify the controlled retest plus original-case checks needed before closure.
Expected outcome
A backup explicitly accepts the case. Closure needs an observed recovery result and original-case reconciliation, not a cleared badge.
Prepared November 8, 2026 by Kashif Rehman. Original planning examples and test records; no customer results or automatic system verification are claimed.
Product information checked November 8, 2026. Confirm current settings and account-specific terms.