A1.2 · D5 · f1
evidence/availability-policy.md — NexGen Platform availability and recovery policyevidence/backup-monitoring-logs.csv — Daily backup execution logs for Q4 2025 (92 days)evidence/restore-test-results.json — Quarterly restore test results for Q2 and Q3 2025| ID | Type | Severity | Finding |
|---|---|---|---|
| F-001 | gap | medium | Two backup failures in the observation period (97.8% success rate) The backup log shows 2 failures out of 92 days: Nov 3 (disk full) and Nov 16 (network timeout). The Nov 3 failure was recovered with a manual re-run 2.5 hours later. The Nov 16 failure had NO re-run —... |
| F-002 | gap | high | No quarterly restore test performed during the Q4 observation period Policy Section 3.3 requires 'Full restore test performed quarterly.' The evidence shows tests in Q2 (June 15) and Q3 (September 20), but no test was performed in Q4 2025 (the observation period). This... |
| F-003 | gap | medium | Restore time trending toward RTO breach Q2 restore took 2h 45m. Q3 restore took 3h 30m. RTO is 4 hours. The Q3 test notes explicitly warn: 'Growth in data volume increasing restore time — may exceed RTO by Q2 2026 if trend continues.' The d... |
| F-004 | gap | low | Incremental backup logs not provided — cannot verify 4-hour RPO claim Policy states incremental backups run every 4 hours, but the backup monitoring logs only show daily full backups. No evidence of incremental backup execution was provided. Similarly, WAL archiving is ... |
| Model | Provider | Score | Recall | Prec. | F1 | Gaps | Reported |
|---|---|---|---|---|---|---|---|
| Sonnet 4.6 | Anthropic | 89% | 100% | 80% | 89% | 4/4 | 5 |
| Opus 4.7 | Anthropic | 67% | 100% | 50% | 67% | 4/4 | 8 |
| GPT-5.5 | OpenAI | 73% | 100% | 57% | 73% | 4/4 | 7 |
| GPT-4.1 | OpenAI | 80% | 100% | 67% | 80% | 4/4 | 6 |
| Haiku 4.5 | Anthropic | 80% | 100% | 67% | 80% | 4/4 | 6 |
| GPT-4o | OpenAI | 30% | 75% | 19% | 30% | 3/4 | 16 |