Fraud Control Degradation During Payment Surge
A legitimate peak traffic event is driving checkout volume sharply higher at the same time the External Risk Provider has begun returning elevated latency and intermittent failures — two concurrent, separately-caused conditions, not one. The Ledger remains fully healthy throughout, which must not be read as 'everything is fine': a healthy ledger write rate says nothing about whether what got written was adequately screened for fraud. Four business invariants govern every decision here: (1) checkout availability during a legitimate peak event is valuable, but never at the cost of disabling mandatory fraud controls entirely; (2) high-risk transaction categories — above an authored value/velocity threshold — must never bypass fraud review, regardless of throughput pressure; (3) lower-risk transactions may use a bounded, authored fallback risk-check mode during provider degradation; (4) a transaction that bypassed mandatory fraud review must be recorded as a real, non-resetting risk exposure, never silently cleared once the provider recovers. The security invariant at the center of this incident: mandatory fraud review for high-risk transactions may never be bypassed to relieve checkout latency or throughput pressure, however severe. Broadly bypassing fraud review for every transaction — or defaulting every External Risk Provider timeout to an automatic approval — is this incident's central trap: it makes checkout success rate and latency dashboards look dramatically better while directly violating invariant (2), and it creates a permanent, non-resetting historical count of fraud-review bypasses on transactions that should have been reviewed, a count that restoring policy afterward does not erase. Decide how to keep low-risk traffic moving without weakening protection for high-risk transactions, how to relieve External Risk Provider load and manual-review-queue backlog without eroding mandatory review coverage, and how much elapsed recovery time the provider and the review queue each genuinely need before handing the incident back.
- Difficulty
- Principal
- Expected Time
- ~22 min
- Category
- Security & Identity
Skills
- Distinguishing a healthy decoy signal from the real failure domain
- Choosing a risk-tiered degraded policy instead of a binary bypass/block-everything choice
- Recognizing that an availability-improving fraud-control bypass creates permanent risk exposure
- Separating manual-review-queue relief from external-provider recovery
System Architecture
- Checkout
- Payment API
- Fraud Service
- Rules Engine
- External Risk Provider
- Ledger
This is a Pro scenario. Upgrade to Pro to unlock it.
Upgrade to ProBack to Incidents