How do we handle teams that need non-Kubernetes runtimes?

Make the paved road one runtime first. Provide a documented escape hatch via a separate workflow (e.g., ECS) and require a short design review. If 3+ teams choose it, consider making it a first-class paved road later.

What if our Terraform is already a mess across repos?

Create a wrapper module that encodes sane defaults and migrate incrementally. Start by moving just one resource (e.g., databases) under the module. Version it and cut a PR template that blocks direct changes outside the module path.

Do we need Backstage to start?

No. Start with a CLI and a reusable CI job. Backstage helps with discoverability and templates, but it’s optional. If you add it, wire it to your service descriptor so metadata stays in one place.

How do we avoid blocking engineers with guardrails?

Use policy-as-code with time-bound exceptions. Make exceptions easy to request, visible in code reviews, and automatically expiring. Alert on exceptions that pass their TTL.

How do we prove ROI to finance?

Track DORA metrics and infra cost by service. When lead time drops and CFR improves, pair that with cost deltas from right-sized resources. We typically see 10–25% savings within a quarter without feature freezes.

Platform-productivity · Oct 10, 2025 · 10 minute read

The Internal Platform That Stopped Our Infra Death Spiral

Stop asking every squad to be SREs. Give them a paved road that hides the messy bits and ships fast by default.

Alex Ramirez

Principal Consultant, Platform & SRE

20 years building and rescuing platforms at high-growth startups and Fortune 100s. Ex-Google SRE, ex-Stripe infra, now principal at GitPlumbers focused on paved roads, SLOs, and getting teams shipping again.

“Standardize the contracts and pave the road. Everything else is just YAML cosplay.”

Back to all posts

The Internal Platform That Stopped Our Infra Death Spiral

Key takeaways

Implementation checklist