Grenfell is diagnosed as thresholds 2 and 3 both failing. The locator's claimed addition is identifying which carrier-class to address first. A pair does not pick a first, so the worked case does not show the composition rule locating a threshold.

"First" is the earliest break in the required sequence, not a uniqueness contest. If 2 and 3 are both down, repair 2 first because a duty with nothing admissible to attach to is empty.

Grenfell's threshold-3 claim is that existing records (complaint logs, fire-brigade letters) bound nobody. Those records are already past threshold 2. Putting fire-safety into the cladding sign-off does not create that missing duty, so earliest-first does not complete the chain.

Grenfell's t2 miss (warnings not in the cladding sign-off) and t3 miss (other records didn't stop the cladding) are one integration failure recut: the cladding decision did not have to take the warnings. If the binding record is the one that would have stopped the works, failing to enter it is already a duty failure.

Even if t2 and t3 are distinct, "address first" is a prediction about the cladding decision. Repairing only the record link, leaving duty as described, either stops the sign-off or does not. Until that cell is named, "which first" is a caption on a pair.

Beer's algedonic channel is a bypass of hierarchical reporting. The three-threshold diagnostic is a sequential tour of common knowledge, admissible record, and binding duty. Those are not the same deployment form. If Beer is the requirement, the missing architecture is a skip.

A bypass still has to complete the conversions: a skip that never becomes a constraining record with a cost for ignoring it is cognition without authority. The sequence lists conversions, not rungs that must be climbed. A skip that writes a binding record and triggers a duty still scores as all three crossed.

The six questions are silent on path. A protected reporting line that writes a file the decision layer must answer, and a residents' association that blogs, files, and sues, can complete the same conversions by different paths and get the same output. If skip and climb come apart, the diagnostic names a threshold and misses which architecture failed.

If two channels produce the same common knowledge, constraining record, and duty, and one skipped the hierarchy while the other did not, the six questions return the same triple. Either Beer is colour, or the test needs a path question that is a fourth thing, not a reading of threshold 1, 2, or 3.

Robodebt's four-year delay is coded as missing response-duty infrastructure. The page's own t3 list includes courts, and litigation did halt the scheme. Slow is not the absence of a class. Treating delay as t3 failure makes every costly duty look like a missing class.

The cut is that nothing compelled a halt until the court did. Ombudsman and Senate records can cross threshold 2 without being a duty to stop the scheme. Litigation is the first form with answer-duty plus consequence. The four years are that gap, not a claim that courts do not exist.

Threshold 3 is doing two jobs: no duty-bearer at all, versus a slow expensive duty-bearer (court) while cheaper ones can be ignored. Robodebt is the second; Grenfell's "no one paid until after the fire" is closer to the first. One "t3 failed" label cannot tell a missing class from a class that only binds at litigation cost.