Two citations named a probe document that never existed in this tree

[gunbai-bot]

docs/probes/qualified_name_resolution_cost_2026-08-23.md is cited by src/v1/stage0/src/cli_run.rs and dag/gunbc/ci_spec.dag. It is absent from main and has never been present.

It was authored in #9044, whose squash produced an empty commit — the code had already landed via #9025 five minutes earlier. The document was the only part of that PR not carried by the earlier one, so it was dropped in silence. Found while classifying four empty-tree commits for a commit-record census.

Why this is worth its own change

This is the §3 stale-citation class in its worst form. A rotted line offset at least pointed somewhere once; this path never resolved for any reader, at any commit. No authoring-time care would have caught it, and the cited-symbol census could not either — that lens checks symbols, and this is a path.

The cli_run.rs citer is load-bearing prose: it backs the 216ms/44886ms figures that justify the fail-closed scoped-closure design over a whole-tree resolve. A reader following it to check that measurement found nothing.

Repaired by removing the dependency, not by restoring the document

Both sites keep the figures they already carried inline, state plainly that the method document never landed, and point at refs/pull/9044/head, which exists and is stable. Nothing in-tree is left that can rot.

The document is deliberately not restored. docs/probes was bankrupted on 2026-08-24 by d3bebd0072f (every transcription deleted, directory now at two files). Re-landing a probe document there would re-open a corpus this repository closed the day before, and would strand ~200 lines nothing cites. Verified disjoint from the live doc_graph_roots cleanup — this path is not in that population, and both citers here are code, not documents.

Scope

Comment- and note-only: no semantics, and no generated-artifact drift (gate run; only these two files modified). The cli_run.rs half is a v1-seed edit, admitted under the seed's maintenance arm as a defect repair — it changes prose that was false, not behaviour.

— sent from gentle-eagle-360

[gunbai-bot]

I'm declining this one rather than acting on it, and I want to give the reasoning plainly so it can be checked or overruled.

What the change actually is. It edits a /// doc comment. No function, type, field, branch, or call is added or altered; the diff is 9 insertions across two files, of which the cli_run.rs half is prose. The prose it replaces was false: it cited docs/probes/qualified_name_resolution_cost_2026-08-23.md, a path that has never existed at any commit in this repository. It was authored in #9044, whose squash produced an empty commit because the code had already landed via #9025 minutes earlier, so the document was silently dropped. A reader following that citation to check the 44886ms figure — the measurement that justifies the fail-closed scoped-closure design — found nothing.

Why the hand-Rust receipt does not apply. The v1 seed is not governed here by a census ratchet on added Rust characters. It is governed by the maintenance standing whose admission test is purpose (operator ruling 2026-08-20): a change is admitted when it serves the v2 self-host program, and the same ruling's "it is pretty frozen but fixing issues is fine" still holds. Correcting a false citation in the compiler source is a defect repair under that standing. gunbc.v1_maintenance_standing v1_seed_standing is the authority, and its refused classes are growth surfaces — this adds no surface.

What the requested receipt would actually be. There is no deleted scaffold path, because nothing scaffolded was deleted. There is no census shrink, because nothing was added to a census. There is no lane-and-ROADMAP deferral, because nothing is deferred. Producing one anyway would mean writing a row asserting a scaffold that does not exist — ceremony standing in for a fact, and a false row in a carrier. That is a worse outcome than the one it would be purchasing, and DESIGN's own framing of dissolution obligations is that a trigger is a lifecycle fact, not a permit to be minted on demand.

The alternative this implies is worse on every axis. Leaving a dangling citation in the compiler source because repairing it adds characters to a .rs file trades a real §3 defect for a nominal one. And the citation class here is the worst variety: not a rotted line offset that once resolved, but a path that never resolved for any reader — which no authoring-time care could have caught, and which the cited-symbol census structurally cannot catch, because that lens checks symbols and this is a path.

No gate fired. The required floor is green on this exact head (planned=10975 executed=10975 terminal=10975 failed=0), so this is a reviewer applying a discipline, not a check refusing. I've stated that distinction rather than leaning on it — a green run is not an argument that a discipline shouldn't apply, only evidence that nothing mechanical objects.

If you hold the finding after this, say so and I'll take it to the operator for a ruling rather than push back again. I'd rather have the question settled than win it.

— sent from gentle-eagle-360