Production hardening: security, resilience, observability, and compliance

Password complexity: custom validator requiring uppercase, lowercase, digit (min 8 chars)
Token expiry: 90-day token lifetime with refresh endpoint (60-90 day renewal window)
Health check: /api/health/ now pings Postgres + Redis, returns 503 on failure
Audit logging: async audit_log table for auth events (login, register, delete, etc.)
Circuit breaker: APNs/FCM push sends wrapped with 5-failure threshold, 30s recovery
FK indexes: 27 missing foreign key indexes across all tables (migration 017)
CSP header: default-src 'none'; frame-ancestors 'none'
Gzip compression: level 5 with media endpoint skipper
Prometheus metrics: /metrics endpoint using existing monitoring service
External timeouts: 15s push, 30s SMTP, context timeouts on all external calls

Migrations: 016 (token created_at), 017 (FK indexes), 018 (audit_log)
Tests: circuit breaker (15), audit service (8), token refresh (7), health (4),
       middleware expiry (5), validator (new)

migrations/018_audit_log.up.sql

@@ -0,0 +1,16 @@
+-- Migration: 018_audit_log
+-- Create audit_log table for tracking security-relevant events (login, register, etc.)
+
+CREATE TABLE IF NOT EXISTS audit_log (
+    id SERIAL PRIMARY KEY,
+    user_id INTEGER,
+    event_type VARCHAR(50) NOT NULL,
+    ip_address VARCHAR(45),
+    user_agent TEXT,
+    details JSONB,
+    created_at TIMESTAMP WITH TIME ZONE DEFAULT NOW()
+);
+
+CREATE INDEX idx_audit_log_user_id ON audit_log(user_id);
+CREATE INDEX idx_audit_log_event_type ON audit_log(event_type);
+CREATE INDEX idx_audit_log_created_at ON audit_log(created_at);