**Concerns:** This is a payment-processor switch, which is the highest-stakes change category. Even with a feature flag, the risk is moderate-to-high because a misconfigured flag could route customers to a processor that fails for them silently. Recommend rolling out to a small cohort first (5-10% of European traffic) and monitoring for an hour before full European rollout. **Pre-deploy:** 1. **Are Adyen API credentials configured in production?** Why this matters: payment integrations fail invisibly if creds are wrong; this is the most common gotcha. What "no" means: stop and configure first. 2. **Is the feature flag currently OFF for all customers?** Why: rolling out code with the flag already ON would route customers immediately, eliminating the gradual-rollout option. What "no" means: stop and turn off the flag, then deploy. 3. **Has the Adyen integration been exercised end-to-end in staging with a real (sandbox) payment?** Why: unit tests catch logic errors; only end-to-end testing catches integration errors. What "no" means: stop and run the staging test. 4. **Is there a designated person to monitor the rollout for the first hour?** Why: payment failures need human response, not just a Slack alert. What "no" means: postpone until coverage is arranged. 5. **Is the cohort selection criterion correct?** (Likely: customer billing country = a European country code) Why: a typo in the country list could route US customers to Adyen or fail to route European customers. What "no" means: fix and verify the criterion. **During deploy:** - Deploy logs from your CI tool (check for build or push failures). - After deploy, the service-health endpoint returns 200 in the target environment. **Post-deploy verification (next 5-30 minutes, with flag still OFF):** - Service is up and healthy. - Adyen API connectivity check: hit the /diagnostics/adyen-ping endpoint (or equivalent) to verify creds work. - No regression in Stripe-based flows: place one test transaction through the existing Stripe path and confirm it works. **Then, flip the flag for the pilot cohort (5-10% of European customers) and watch for 60 minutes:** - Payment success rate for the pilot cohort vs. baseline (Stripe European). Acceptable: within ~1% of baseline. Concerning: drop of 3% or more. - Specific Adyen error codes in logs: anything in the "card declined for issuer reasons" range is normal noise; anything 5xx from Adyen's side is not. - Customer support inbox: any new tickets mentioning "payment failed" or "credit card not working" from European customers. **Rollback trigger:** - Pilot success rate drops by 3% or more relative to baseline, sustained for 10+ minutes. - Any 5xx response rate from Adyen above 1%. - Customer support reports two or more independent payment-failed tickets from pilot cohort. Rollback action: flip the feature flag to OFF. Code path reverts to Stripe immediately. No deploy revert needed unless the deploy itself broke something independent of the flag.