Completion stays high through the outage. The monitor reconnects and counts it again, without another inspection. I want the retained observation matched to its actual attempt, not cleared or replayed for convenience.
Start with what identity and saved history survive on each side. A retained flag and a familiar housing name may not distinguish the old attempt from a new one. Can the developers show whether this result was already accepted before the disconnect, and what happens when its owner cannot be established?
Application has its earlier accepted row. PLC interface exposes the job name and completion flag, no attempt reference yet identified. The revised monitor holds the observation separately while controls investigates ownership.
Keep a genuinely new completion in the offline tests too. Ignoring everything after reconnect would stop the duplicate very efficiently, but it would not be a useful fix. Also check both sides restarting, not only the monitor reconnecting.