Skip to content

Error-to-action matrix

Partner-safe mapping from common failures to caller action, support action, engineering escalation, and retry. Complements the error catalogue.

Legend

Column Meaning
Caller What the integrator / client backend should do
Support What MasteryHive support does with triage evidence
Eng escalate? When to page engineering
Retry? Safe to retry same logical request?

Fraud

Symptom / code Caller Support Eng? Retry?
401 wrong scheme / bad JWT Fix ApiKey + claims/keys — auth tutorial Confirm key exchange env If widespread after rotation After new token
422 missing x-tenant-id Send header — No After fix
Scoring 5xx / timeout Backoff; correlate transaction_id Check service health (internal) Persistent Cautious yes
ML_SCORE_TIER_CONFLICT / AMBIGUOUS_PREDICTION_CONTRACT Do not retry same body; fix emitter / await policy alignment Log as contract defect Yes (known handoff class) No same body
200 score, no report Follow missing report Verify callback reachability with eng If callback healthy but silent Rescore only after cause fixed
Case-rule 400/500 Fix payload; see Case management Correlate Persistent After fix

KYC / KYB

Symptom Caller Support Eng? Retry?
Gateway auth failure Fix env credentials Ingress/provisioning Misprovisioned env After fix
5xx / timeout Treat as uncertain; backoff Vendor/status check Vendor outage Cautious; avoid duplicate billed calls
200 no-match Show as negative verification — Policy only No
200 uncertain / partial vendor fields Show uncertain — do not coerce pass — Rare No
Provider outage pattern Fail closed or queue per your policy Follow internal outage runbook Yes if confirmed After provider recovery

Reconciliation

Symptom Caller Support Eng? Retry?
409 active session / period lock Wait or new session_id Check stuck lock (internal) Stuck lock New session preferred
Ingest 422 Fix scope/template/enums Schema drift check If OpenAPI wrong After fix
Status not terminal Poll status; avoid same-session hammer Stuck-session runbook (internal) Yes if stalled Prefer guidance from support
Resolve without resolution_request_id (messaging) Always send stable id — No After adding id
Duplicate trigger no-op Use new session if intentional rerun Explain trigger guard No New session

Case management (rule scoring only)

Symptom Caller Support Eng? Retry?
Expecting durable case desk from AI score Read Case Management product Clarify scope No N/A
401/422 on case-rule route Same as Fraud scoring auth — No After fix