Two users edit the same record. How would your full-stack feature avoid silently losing changes?

In production I would track conflict rates because frequent conflicts can mean the workflow needs locking, assignment, drafts, or better collaboration design. Some resources need pessimistic locking or a checkout model, but that adds stale locks and support work. I would ensure audit history records both successful changes and rejected stale attempts where useful. API clients need clear status codes and error shapes so the UI never treats a concurrency conflict as a generic failure. I would test the same conflict through the API and UI so the backend protects the record and the frontend gives a clear recovery path.