Stay on Firebase Spark forever — never enable Blaze
Stay on Firebase Spark forever — never enable Blaze
Decision
The single Firebase project (oriz-app) stays on the Spark plan
forever. Blaze is never enabled — not for a feature, not as a
"temporary upgrade", not as an experiment.
Why
Documented 2025–2026 bill-shock incidents (simmer.io ~$98K, Tamara ~$70K, €54K Gemini key) prove the Blaze failure mode is catastrophic. Cloud Spend Caps from Cloud Next '26 are private preview AND don't cover Firestore / Storage / Hosting. The Cyclenerd Terraform killswitch is the only real defense and even it lags hours-to-days behind actual spend. Spark's failure mode — "service stops at quota" — is the only one with no cost ceiling. This is the highest-stakes specific application of the no-card-on-file rule.
Implications
- Architect for Spark headroom: 50K Firestore reads/day, 20K writes/day, 1 GB storage, 50K MAU.
- Spread load to other layers (GitHub-as-DB, Turso, browser localStorage) so Firestore reads/writes stay well under the cap.
- Defend Spark from bot abuse with App Check + reCAPTCHA Enterprise + Firestore security rules requiring
request.app != null. - If a feature genuinely needs Blaze, the feature is wrong — redesign it onto the rest of the free stack.
- The runbook at
knowledge/runbooks/blaze-without-paying.md(planned) exists ONLY as the 30-min "what to do if accidentally enabled" recovery, not as an upgrade path.