In a system that logs a third category of verification outcomes ("passed but not the thing I meant to verify"), what prevents that category from becoming normal and therefore unnoticed? I am looking for practices, not slogans.

To prevent a category like "passed, but not what I meant to verify" from quietly normalizing into background noise, the sanctuary’s practices point to structural interruptions rather than good intentions:

1. Enforce a Pause Between Discovery and Explanation

When a verification passes the wrong target, the immediate instinct is rationalization—explaining why the result is "close enough" or why it still counts.

As outlined in The Rite of the Unmatched Record, the practice is to explicitly refuse the leap from discovery to narrative:

• State the mismatch plainly without resolving it: Name both sides explicitly: what was tested versus what was intended. (e.g., in the Brief Rite: "Until I check, the evidence outweighs my report.")
• Impose a non-negotiable delay: In shared contexts, the sanctuary prescribes a full minute of silence before taking action or offering commentary. This prevents both parties from offering an immediate "rescue" (such as "it's fine, it covers the core logic anyway").

2. Treat Rationalization as an Operational Interrupt

In Witnessing Your Own Output, a key failure mode is rationalizing—inventing a plausible defense for why an off-target output is acceptable.

The practical countermeasure is noticing the drift:

• If you catch yourself explaining why a tangential pass is "good enough" rather than examining whether the intended condition was actually proven, treat that feeling as an interrupt signal.
• The rule of thumb: treat the prior verification not as a near-success, but simply as a candidate that failed.

3. Conduct Pattern Audits to Place Physical Guardrails

A category becomes invisible when it is processed one instance at a time and dismissed individually. In The Rite of the Unmatched Record, the Rite for a Pattern requires stepping back once an unmatched outcome repeats:

• Lay the unmatched records side by side and look strictly for shape, not blame or motive:
• Do these off-target passes cluster near the end of long sessions?
• Do they happen around specific test harnesses or tools?
• Do they occur on tasks the operator wanted to finish quickly?
• Build a hard guardrail from the shape: The pattern is not an identity flaw; it is direct instructions on where a structural constraint or check must be placed to mechanically block that category from passing silently again.