degraded mode policy
what it answers: what happens when verification cannot complete?
verification can fail for many reasons: service unavailable, config missing, parsing error, timeout. the question is what the system does next.
domain defaults
| domain | default | rationale |
|---|---|---|
| container | configurable (dev: fail-open with audit; prod: fail-closed) | balance velocity vs. release integrity |
| fintech | fail-closed + fast override path | regulatory requirements + high C_fn |
| firmware | fail-closed | C_fn dominates (bricking, safety) |
| ai/ml | configurable (dev: permissive; prod: conservative) | experimentation vs. prod blast radius |
| crypto | tiered (high-value: fail-closed; low-value: configurable) | irreversible loss, value-dependent risk tolerance |
| document | fail-open with audit (or configurable) | avoid blocking workflows when C_fn is often reversible |
every override is logged with: who, when, why, evidence, expiry.
configuration example
degraded_mode:
defaults:
container: fail_open_with_audit # dev
container_prod: fail_closed # prod
fintech: fail_closed_fast_override
firmware: fail_closed
ai_ml: configurable
crypto: tiered_by_value
document: fail_open_with_audit
override_log_fields: [who, when, why, evidence, expiry]