Fixes
- Provider pools now restore a small validated connection reserve promptly after startup or idle expiry, keeping the first preparation and health check ready without waiting for a full warmup cycle.
- Health checks keep recovery capacity available for plain Backup providers even when health-capable providers fill the global connection limit. Backup connections are used only for unresolved articles and released afterward for normal work.
- Failed Watchdog attempts now expose the actual failure reason, the stage that stopped, partial preparation and health timings, and any provider activity captured before the failure.