Loading service level objectives…
The error budgets BurnRateSurgeon holds itself to, measured by BurnRateSurgeon: API availability and latency, replay completion and replay latency, over a rolling 30-day window.
An objective is outside its error budget.
Requests to the API that this service answered rather than failed on. A refusal the caller can act on — an invalid body, a rejected credential, a spent rate-limit allowance — is a correct answer and counts as good; a 5xx does not.
935 events measured · last evaluated
Requests answered within 200 ms, measured from the moment the Worker was entered — before the CORS allowlist and before the credential gate (CONST-DOD-06, ADR-0017).
935 events measured · last evaluated
Replay jobs that reached a verdict. A job that exhausted its retries or could not be graded has failed the customer even when every HTTP request around it answered 200.
no events recorded yet · last evaluated
Replay jobs graded within 60 s, which is the budget ADR-0018 sets for the largest window the product sells.
no events recorded yet · last evaluated
Every number above is measured by this deployment about itself and graded once an hour against a rolling 30-day error budget, on the same multiwindow burn-rate arithmetic the replay engine applies to a customer's own rules. Nothing here is per-customer.