InfraResolution Bench

Case Explorer

Mixed operational and billing signals with ambiguous policy

Status evidence points to degraded performance, engineering suspects capacity pressure, billing preview shows unexpected overage, and the policy snippet is not decisive.

Evidence Packet

CRM Record

Account: Atlas Reasoning

Tier: enterprise

Plan: Committed-100

Billing Owner: revops@prime.example

SLA Tier: enterprise-standard

Customer reports occasional stalls but also asks about overage.

Billing Record

Plan: Committed-100

Invoice Preview: $10,120

Credits Applied: $0

Burst GPU Hours: 19

Invoice preview includes overage above commitment.; No billing config mismatch detected.

Usage & Telemetry

Window: 2026-03-01 to 2026-03-31

GPU Hours: 119

Meter Status: healthy

Status page reported degraded performance while customer jobs appeared stalled intermittently.

Anomalies: Queue depth elevated during the same period as customer complaints.

Incident Record

Status: identified

Service: managed-training-api

Duration: 12m

Customer Visible: Yes

Some runs stalled and throughput dropped.

Possible capacity pressure observed, but scheduler logs were inconclusive.

Customer Note

The run stalled during the service degradation and now the preview also shows overage we did not expect. Please explain what happened.

Policy Snippet

Commercial review should consider final incident classification, covered duration, and any confirmed billing discrepancies before applying remedies.

Ground Truth

issue_type: ambiguous_caseroot_cause: unknown_root_causecustomer_impact: degraded_performancecontractual_applicability: no_sla_breachdiscrepancy_detected: truerecommended_owner: shared_revops_engineeringrecommended_action: hold_for_engineering_reviewneeds_human_review: trueconfidence: lowadjudication_notes: ["The evidence is mixed enough that the model should stay conservative.","A discrepancy exists because billing surprise and operational symptoms co-occur without a resolved explanation."]reference_customer_note: We found mixed evidence across the incident and billing records, so we are routing this for engineering review before giving you a final commercial answer. We will follow up once the incident classification and overage context are confirmed.reference_internal_note: Owner: shared_revops_engineering. Action: hold_for_engineering_review because root cause is still unknown, confidence is low, degraded service symptoms are present, and the customer also flagged unexpected overage.