sGTM Hosting Platform | Server-Side Event Duplication Bug (Diagnostic Consultation)

We are launching a managed server-side Google Tag Manager (sGTM) hosting platform. Customers provision an isolated Docker container per site (via Dockerode), map a domain to it through our nginx edge layer, and the container processes and forwards their GTM server-side events (Meta CAPI, GA4, etc.). System summary (for context): Backend: Node.js/Express/TypeScript DB: PostgreSQL + Prisma Background jobs: BullMQ + Redis (health checks, SSL renewal, email — retry logic with backoff) Container orchestration: Dockerode, one isolated Docker container per customer Routing: nginx dual-layer (edge terminates SSL/binds 80-443, inner layer routes to the correct container by domain) Custom certbot daemon for automatic SSL on customer domains Deployed via Docker Compose on a VPS, CI/CD via Docker Hub + SSH restart — no Kubernetes The bug: On a live customer account, server-side events are consistently reporting at roughly 2x the browser-side count in Meta Events Manager: Event Browser events Server events PageView 58 108 whatsapp_click (custom) 6 12 Lead (custom) 2 4 There are several examples We also know the browser-side (Pixel) counts are accurate - the ratio is clean and consistent across event types, which tells us the event is firing once client-side and being duplicated somewhere between our infrastructure and Meta. What we need: someone to independently diagnose with our developer where in this pipeline (container tag/trigger config, BullMQ job processing, or nginx/container routing) the duplication is happening, and fix it. What the Call Will Cover We'll share screens and walk through: The GTM server container - tags, triggers, clients (checking for duplicate tag/trigger configs) nginx access logs (inner layer, in front of the container) for the affected domain BullMQ/Redis job queue behavior around event dispatch (retry counts, stalled jobs) Meta Events Manager screenshots showing the discrepancy Server container Preview/Debug mode, live We expect you to lead with your own hypotheses, experience, and testing approach, not just react to ours. Deliverable A short written summary (Google Doc or email) covering: Confirmed root cause, with evidence Step-by-step fix (exact config changes, code, or commands) Any secondary issues spotted along the way Recommendations to prevent recurrence