v2.0.8
Reliability fixes for the websocket connection and CT sensor data validation. No config or entity changes — safe to update in place.
Fixed
- The integration could freeze for hours with no errors logged, requiring a manual restart (and sometimes reauthentication) to recover. The 60-second "no data → resubscribe → force reconnect after 5 tries" safety net reset its staleness clock on any text frame received from the cloud, before even checking if it carried real data. If the cloud connection ever degraded to sending only non-substantive traffic, the safety net would never trip and the connection would look "alive" indefinitely while delivering nothing. It now only resets the clock on a genuine data notification.
- Silent auth failures during websocket keepalive. The periodic bandwidth-keepalive PUT and API-version heartbeat GET calls logged their HTTP response status at debug level unconditionally — an expired auth token during a stall would never have surfaced anywhere visible, and nothing would react to it. Both now log a clear warning on 401/403/406 responses.
- CT power spike guard had no absolute ceiling. The existing guard only required two similarly-sized bad readings in a row to accept an obviously-impossible value (e.g. a spurious ~-374,000 W reading on a residential CT) into sensor history. Added a hard ceiling (50,000 W) that rejects outright with no confirmation path — no real reading should ever come close to it, so this can't cause false rejections of legitimate spikes.
Internal
- Extended the websocket keepalive logging so 401/403/406 responses are distinguishable from ordinary transient failures in the log, aiding future diagnosis.