API Failure Mode

A practical example

Example: a vendor deprecates a field and renames it. The integration keeps returning 200 responses, but every sync silently writes nulls.

That is a schema-drift failure mode, and it needs contract testing, not just error alerts.

What to evaluate before investing

• Ask vendors to document their rate-limit behavior: what happens at the limit, and how is backoff signaled?
• Check whether the platform detects schema or field changes in connected systems, or only HTTP-level errors.
• Review what fallback each failure mode triggers: retry, queue, dead-letter storage, or silent drop.

Limitations and tradeoffs

Failure modes are pattern categories, not a formal standard; naming and handling differ by vendor, so map each mode to a concrete response during evaluation.