Reliability and monitoring

The network now detects failures earlier and gives the team enough context to fix them quickly.

• Health probes continuously cover the Minecraft server, website, PostgreSQL, and RabbitMQ.
• Prometheus, Grafana, Loki, and Alertmanager track usage, performance, logs, and fatal conditions.
• Critical failures notify the Cookie Build team through Discord without interrupting gameplay services.
• Session cleanup and match accounting now guard against stale or duplicated player state.