Error handling strategies in OIC — what does your production setup look like?

Error handling is where most OIC integrations are under-engineered.

Layer 1: Retry scope on all REST adapters — 3 retries, exponential backoff (30s, 60s, 120s). Retry only on HTTP 5xx, never on 4xx.

Layer 2: Global fault handler that writes to ATP tracking table and sends email notification.

Layer 3: Weekly Error Hospital review — any pattern that appears 3+ times gets a root cause fix.

The key insight: separate "expected errors" (business validation failures) from "unexpected errors" (network timeouts, API downtime). They need completely different handling.