Connection pool metrics are monitored

Connection pool exhaustion produces silent request failures — requests queue behind a saturated pool, timeout, and return errors — with no visibility into why. CWE-400 (uncontrolled resource consumption) applies when pool limits are reached without detection. Without pool metrics, the team discovers saturation only when users report errors, not proactively when the pool approaches its limit. ISO 25010 resource-utilisation requires that resource consumption is observable; connection pool health is a critical operational signal for scaling decisions and for diagnosing connection leaks introduced by code changes.

Add pool event listeners to log connection activity and expose pool metrics through the /api/health endpoint. For Prisma, use the built-in $metrics API.

// pg Pool — monitor pool events
pool.on('connect', () => {
  console.debug(`[pool] +1 connection. total=${pool.totalCount} idle=${pool.idleCount}`)
})
pool.on('acquire', () => {
  if (pool.waitingCount > 0) {
    console.warn(`[pool] ${pool.waitingCount} requests waiting — consider increasing pool max`)
  }
})
pool.on('error', (err) => {
  console.error('[pool] Idle client error:', err)
})

// Expose in /api/health response:
// { pool: { total: pool.totalCount, idle: pool.idleCount, waiting: pool.waitingCount } }

// Prisma metrics (4.9+)
const m = await prisma.$metrics.json()
// m.gauges: pool_active_connections, pool_idle_connections, pool_wait_count
// Forward to Datadog/Prometheus/your metrics backend