Infrastructure · Verification

DetectorSelfTest

Test the detectors that guard a process, so their silence is not mistaken for health

Definition

Mechanism

On a stated schedule, known faults are injected or replayed, and each Monitor or check that guards a process is confirmed to catch them. A detector that misses a known fault is marked failed and repaired before the process relies on it again.

Invariants

  1. Each detector is tested with a known fault on a stated schedule.
  2. A detector that misses a known fault is marked failed and is not relied on until it is repaired.

Risks

Failure modes

  1. Silent monitor

    a detector stops firing and its silence is read as health.

  2. Test-only detection

    detectors catch the injected faults but not real ones.

Caution

Injected or replayed faults can fire the real responses their detectors drive — keep test faults tagged and isolated from live state.

Relationships

Builds on

References

Monitor

Continuous observation of state over time

Used by

References

ClosureConservationCheck

Check that a conserved quantity balances over every holder, as an equality, not as a lower bound on the holders being watched

Mind

MeaningChangeSweep

When a value's meaning changes, find and recheck every reader of the old meaning

Reference

Identity

Ref
DetectorSelfTest#8d48
Sema ID
sema:DetectorSelfTest#mh:SHA-256:8d48b9e7f6dbfc6ea93f0de86259339a64d1ad994089f2ff5d8642b1d2cc842a
Ring
2
Tier
2
Updated
2026-11-11