The API runs on a 4 GB RAM VPS, Nginx in front, PHP-FPM configured with 50 workers. A traffic spike — nothing exceptional, a marketing campaign — and in 8 seconds the pool is saturated. CPU at 30%, server down. Monitoring showing 502s and 504s in bursts. The bottleneck wasn't the CPU. It was RAM and the exhausted pool. Six months later, migration to Go. Not out of hype, but because we understood t