IncidentSutra AI Incident evidence report · created by Vatsal Shah · Observe precisely. Reason cautiously. Recover responsibly.
inc_e95e3ae334e5f1390998 HUMAN REVIEW REQUIRED offline · read-only root cause: unconfirmed 0.2.1
SYNTHETIC DATA — this report was generated from fictional demonstration records, not a real incident.
Summary Timeline Hypotheses Gaps Quality & privacy Sources Limits
Summary 9 observed events
3 error/alert signals
3 changes
2 metric events
5 competing hypotheses
1 evidence gaps
Window 2026-11-10T19:40:00.000Z → 2026-11-10T20:02:00.000Z · services: checkout, notifications, payments
observed in an imported recordinferred hypothesis, never confirmedunknown data not suppliednot run deliberately not executed
Auditable timeline (UTC) Event timeline by service 9 observed events between 2026-11-10T19:40:00.000Z and 2026-11-10T20:02:00.000Z. Diamonds are changes, filled circles are error or alert signals, squares are other observations. A table with the same data follows. checkout notifications payments 2026-11-10T19:40:00.000Z metric: Baseline checkout p95 latency 220ms, error rate 0.1% 2026-11-10T19:45:00.000Z deploy: Checkout release 2.7.0 rollout started 2026-11-10T19:49:00.000Z error: Intermittent checkout payment dependency timeout before second deployment 2026-11-10T19:52:00.000Z deploy: Checkout release 2.7.1 rolled out 2026-11-10T19:54:00.000Z alert: Checkout error budget burn above threshold 2026-11-10T19:55:00.000Z metric: Checkout p95 latency 2100ms, error rate 8% 2026-11-10T19:56:00.000Z error: Payment gateway timeout; Bearer [REDACTED] should be hidden; contact [REDACTED_E 2026-11-10T19:57:00.000Z deploy: Unrelated notifications release 2026-11-10T20:02:00.000Z log: External payment processor increased latency 2026-11-10T19:40:00.000Z 2026-11-10T20:02:00.000Z Diamonds: changes (3). Circles: error/alert signals. Squares: other observations.
Time Service Type Severity Observation Evidence 2026-11-10T19:40:00.000Zcheckout metric info Baseline checkout p95 latency 220ms, error rate 0.1% evt_969c8550e1b83b9d6d92observations.jsonl:1 2026-11-10T19:45:00.000Zcheckout deploy info Checkout release 2.7.0 rollout started evt_9a1ec8c4509bf193501eobservations.jsonl:2 2026-11-10T19:49:00.000Zcheckout error error Intermittent checkout payment dependency timeout before second deployment evt_5cc6ce3f88bbe686a2b1observations.jsonl:3 2026-11-10T19:52:00.000Zcheckout deploy info Checkout release 2.7.1 rolled out evt_cc4992c8882161412b0aobservations.jsonl:4 2026-11-10T19:54:00.000Zcheckout alert critical Checkout error budget burn above threshold evt_9a8cc244ee051d5ad97aobservations.jsonl:5 2026-11-10T19:55:00.000Zcheckout metric warning Checkout p95 latency 2100ms, error rate 8% evt_591427918bf4b89155c4observations.jsonl:6 2026-11-10T19:56:00.000Zcheckout error error Payment gateway timeout; Bearer [REDACTED] should be hidden; contact [REDACTED_EMAIL] evt_5ef4f8aad618256ca464observations.jsonl:7 2026-11-10T19:57:00.000Znotifications deploy info Unrelated notifications release evt_4c27901e059d268cdfc4deployments.csv:1 2026-11-10T20:02:00.000Zpayments log warning External payment processor increased latency evt_4db866928afb7dadea76observations.jsonl:8
Competing hypotheses None is confirmed. Each lists what would have to be true, what contradicts it, and read-only checks that would discriminate between them.
Change "checkout-2.7.0" may be related to checkout errors inferred deployment-change unconfirmed confidence: undetermined review: pending_human_review
The change and error signals occur close in time. Timing alone cannot establish causality.
Would need to be true Error rate for checkout rises after 2026-11-10T19:45:00.000Z and not before Cohorts still on the previous version stay healthy Contradicting none Other services changed in the same window without error signals (weakly informative). Would disconfirm Errors found before the change in unsupplied data Rolling back or disabling the change in a canary does not change the error rate (not run) Read-only checks not run Compare error rate and latency for the affected service in matching windows before and after the change (read:metrics, cost low) Review the release diff and feature-flag changes for the affected code path (read:source-control, cost low) Compare an unchanged region, cohort or canary on the prior version (read:metrics, cost low) Change "checkout-2.7.1" may be related to checkout errors inferred deployment-change unconfirmed confidence: undetermined review: pending_human_review
The change and error signals occur close in time. Timing alone cannot establish causality.
Would need to be true Error rate for checkout rises after 2026-11-10T19:52:00.000Z and not before Cohorts still on the previous version stay healthy Contradicting evt_5cc6ce3f88bbe686a2b11 error/alert signal(s) precede this change; the problem pre-dates it. Other services changed in the same window without error signals (weakly informative). Would disconfirm Observed: errors before the change (evt_5cc6ce3f88bbe686a2b1) Rolling back or disabling the change in a canary does not change the error rate (not run) Read-only checks not run Compare error rate and latency for the affected service in matching windows before and after the change (read:metrics, cost low) Review the release diff and feature-flag changes for the affected code path (read:source-control, cost low) Compare an unchanged region, cohort or canary on the prior version (read:metrics, cost low) Dependency or third-party failure may be affecting checkout inferred dependency-or-third-party unconfirmed confidence: undetermined review: pending_human_review
Messages near the onset mention timeouts, upstream, provider or gateway problems.
Would need to be true The dependency degraded at or before onset Failing requests traverse the dependency Would disconfirm Dependency health is normal at onset (not run) Failing traces do not touch the dependency (no trace join run) Read-only checks not run Check upstream dependency health, latency and status notices for the same window (read:metrics, cost low) Join failing requests by trace ID to dependency spans (read:traces, cost medium) Independent service or dependency problem affecting checkout inferred other-or-unmodeled unconfirmed confidence: undetermined review: pending_human_review
Observed errors may come from dependencies, traffic, capacity or prior latent defects rather than a release.
Would need to be true A cause independent of the supplied changes exists Would disconfirm Evidence tying the errors to a specific supplied change Read-only checks not run Check traffic volume, capacity headroom and external provider notices (read:metrics, cost low) Collect affected-service SLO and error-budget data (read:metrics, cost low) Root cause remains undetermined inferred unknown unconfirmed confidence: undetermined review: pending_human_review
The supplied data cannot rule out causes outside the modeled families.
Supporting none
Contradicting none
Read-only checks not run Collect error and alert data and identify affected services and SLO impact (read:metrics, cost low)
Observed metric series Series Points First Last Min Max checkout/error_rate_pct 2 0.1 8 0.1 8 checkout/p95_ms 2 220 2100 220 2100
Evidence gaps unknown Sources use different UTC offsets; all were converted to UTC, but clock skew between sources was not measured.
Ingestion quality & privacy Files 2; rejected records 0; skipped inputs 0; cross-file duplicates dropped 0 UTC offsets: Z×8, +05:30×1 · precisions: s×9 Clock skew: unassessed. Wall-clock ordering across hosts and regions is not proof of causal order; clock skew between sources was not measured. Redactions applied: 2. best-effort masking; not a DLP guarantee; sanitize exports upstream.
Sources File SHA-256 Records deployments.csv 529732004d8da13a70b348e7ae95d315fee0018cab187e0df49f3ad1dd7789ad1 observations.jsonl 31a737fee9d579b9173ffeb3ea2cb37d957b4dce92e54085ff896a85d122be058
Approval status and limitations not run production diagnostics · production mutations · trace joins · baseline comparison against historical windows · calibrated confidence scoring
No mitigation, rollback, diagnostic query or production mutation was executed. Temporal correlation does not establish causality. Redaction is best-effort, not DLP. The report digest detects accidental change only; it is not a signature or chain of custody.
IncidentSutra AI · Apache-2.0 · Created by Vatsal Shah · schema incidentsutra.report.v2