Fixed
- Connection errors now write exactly one
action="error"log row per outage instead of one per retry, preventing the dashboard 24h errors counter from inflating during startup races or service restarts (#140) - A recovery
action="info"row is now written tosearch_logwhen an unreachable instance becomes reachable again, making the recovery event visible on the Logs page (#140) - A 10-second startup grace delay before the first search cycle gives co-located Sonarr/Radarr services time to become ready (#140)