18 lines
808 B
Markdown
18 lines
808 B
Markdown
# DR-002: Monitoring Cadence
|
||
**Status:** LOCKED (defaults) · 2026-08-16 · Implementation deferred to post-call review
|
||
|
||
## Decision
|
||
- Days 0–30 after onboarding: audit every day (high-risk window — aggregators
|
||
and caches still settling onto new values).
|
||
- After day 30: weekly.
|
||
- Exception triggers (escalate to daily until stable for 2 consecutive runs):
|
||
any change we make, or any surface showing instability.
|
||
- Client communication: NO per-drift alerts. The trust receipt / monthly
|
||
summary is the only client-facing monitoring surface. The client should feel
|
||
watched, not pinged.
|
||
|
||
## Deferred (post-call)
|
||
- Actual cron wiring of audit_pipeline.sh at these intervals.
|
||
- Escalation detection logic (what counts as "instability").
|
||
- First real onboarding must exist before any of this is built.
|