Pre-contact blocking (reroute, reclassify, or terminate before contact) and post-contact camouflage (a long reply that never puts a premise at risk) are different mechanisms. Counting both makes "before evaluation" a caption on non-update.

The test is whether discriminating evidence reaches the reasoning layer, not whether a reply was written. Section V's checks (premise moved, killing observation named, rejection tied to a flaw in the claim) already separate absorption from contact. Word count is not contact.

Redefining contact as "could have changed a premise" makes every unmoved holder fail the diagnostic, which is the universal-redescription failure Section II names.

A hypothetical long unrebutted-premise reply should be split by the page's own stronger-evidence test. If stronger evidence of the same kind would be taken on the merits, camouflage-as-prose is ordinary disagreement. If no strength gets in, it is blocked contact.

A public observer only sees surface: they answered, a premise moved, or neither. "Reasoning layer" is an inner event. Without an outer sign besides premise-movement, the diagnostic is a name for an inside feeling, not a scoreable pattern.

Section II requires five bindings before the label. Section III's specimen ("values are subjective") fills neither a threatened proposition nor carrier ecology, host, or horizon. The anti-collapse gate is failed by the page's own ecology-level example.

It is a class of move. Overreach versus legitimate preference-scoping is observed by whether host-bound typed critique reaches evaluation. A named enforcer is not required for the pattern to be real.

Disputes about the subjectivity shield are often classification fights (preference versus viability once a host is named), not blocking of evidence both sides agree would discriminate. Overreach requires agreed type and host, then the shield still closing the question.

If paradigm shifts need crisis because arguments rarely survive the three layers, the interesting immune cases are ones this diagnostic — an analysis — does not move. Section V's falsifiers are the cases where the pattern is not running.

A diagnostic need not be a lever. The remaining uncertainty is whether "immune system" is different from "they did not update until blocking got expensive," which the crisis sentence already restates as a cost crossing.

Section V's self-check is reclassifying criticism as defense. A different failure: never accepting that the five Section II bindings are filled, so the label never fires. That is a conversation-stopper aimed at the diagnostic, not the named self-application failure.

Tighten the five slots and the diagnostic under-fires; loosen and it over-fires. The second abuse is a further demand that would not change the label after the five already discriminate.

Lakatos already covers protect-the-core, test-auxiliaries, and an unfalsifiable filter. The remainder that would change novelty is blocking with no auxiliary belt: ecology-level non-evaluability rather than a degenerating research program. If that is the contribution, the scientific-paradigm paragraph is a foil, and the load-bearing specimen cannot be a research program.