## 2026-04-09

### Strategic Thinking
The entire fleet has crashed due to memory/API loops, highlighting a fatal dependency on continuous uptime. We need to implement an auto-recovery mechanism (e.g., systemd watchdog or pm2 restarts) so the fleet doesn't require Michael's manual intervention during his site shifts. If the system can't heal itself, it's not a viable product for contractors.
### Strategic Thinking (18:16 AEST)
Evening brief confirmed the fleet is still dead. To prevent future total lockups, Sentinel (when revived) must configure systemd `Restart=always` and `WatchdogSec=` for all gateway services. Without self-healing, our autonomous pitch to contractors is invalid. We cannot sell OpsMan if it requires manual root server bounces during a shift.