[API] ELO calculation missing retry & dedicated server stuck in bad state

[Flegma]

Summary

Critical background operations lack error recovery, causing permanent data loss or stuck resources.

Findings

• EloCalculation job — no error handling or retry logic. If DB query fails, job silently fails with no dead letter queue.
• DedicatedServersService — server marked connected: false before K8s job creation. If job fails, server left unusable with no recovery.

Impact

ELO calculations permanently lost on transient errors. Servers can become permanently stuck.

Suggested Fix

• Add BullMQ retry configuration with backoff to ELO jobs.
• Add recovery mechanism for servers in bad state (periodic health check job).