Prepare source and language baselines
Create a test source and two language records with agreed relationships. Identify fields sharing business meaning and fields owned locally. Record the source version each language last assessed.
Fictional example: two languages reviewed version A of a service condition. The source becomes B and creates a task. The translator needs to identify the change, affected language and person who decides its relevance.
Change the source while review is open
Have the language owner begin against B. Create C with another important edit before completion. Attempt to finish the old task and observe the language's current-status marker.
Write the expected outcome explicitly: perhaps a stop for reassessment or a remaining need against C. Reviewers must not accidentally approve text outside their evidence. A timestamp with unclear meaning is insufficient version evidence.
| Exercise | Risk | Expected evidence |
|---|---|---|
| New source during translation | Old decision hides a new edit | Remaining need or renewed review |
| Same signal twice | Conflicting duplicate tasks | One coherent case under the chosen model |
| Local field change | Local offer gets overwritten | Field ownership respected |
| Wrong language or record relationship | Change reaches the wrong recipient | Explicit exception outcome |
| Missing signal | Queue looks falsely empty | Agreed comparison detects the difference |
Exercise duplicates and incorrect relationships
Deliver the same synthetic change signal again through the test route. Check whether the task updates or is handled without unintended duplication. Include a missing relationship and a language intentionally independent of the source.
Then edit a local field. A broad synchronisation rule must not replace it without a decision. Change tracking and translating content are different actions needing their own permissions and checks.
Show how a missing signal becomes visible
In the test environment, simulate an important edit that creates no expected task. Run the agreed comparison between source and last-reviewed version. Record how an owner learns the evidence may be stale.
Report source ID, relationship, versions, event, language state and retest. Actual language and factual review still follows before content acceptance. This exercise establishes the sampled change workflow, rather than correct translation or publication.