← Back to rankings

Synthetic test history

refund_webhook_idempotency

tests.refunds::WebhookContract payments.yml · test (node-22, postgres-17) · ubuntu
Observed nonpass rate · reversal cohorts12.7%
95% Wilson interval · reversal cohorts8.2–19.1%
Conflict groups9
Estimated CI impact / 30d126 min

The evidence

Same-input reversals

Demo mode · synthetic data

Representative rows from the synthetic history; the summary metric includes all 9 conflict groups.

Representative same-input run history for refund_webhook_idempotency
CommitAttemptOutcomeJob durationObserved
dc70a42 1 fail 6m 48s 5h ago
dc70a42 2 pass 6m 36s 5h ago
0ff831a 1 pass 6m 22s 2d ago
0ff831a 2 fail 7m 03s 2d ago
724b9ee 1 fail 6m 51s 6d ago
724b9ee 2 pass 6m 40s 6d ago

Rows are paired because both the commit SHA and execution fingerprint match. Different SHAs or runtime matrices are never compared.

Before you connect

Public repositories only

Quarantine uses GitHub OAuth to identify you and list public repositories. This prototype requests no private-repository or write scope.

Retention and revocation: signing out invalidates active Quarantine sessions. Revoking the OAuth grant in GitHub invalidates the credential but does not delete its encrypted database record. Only operator database cleanup deletes that record; this prototype has no self-service account deletion. Use the synthetic demo if you do not want to connect an account.

For private repositories, the safer production path is a GitHub App with repository-level Actions: read access.