Delete supply-side lens enforcement from floor discovery

DESIGN.md

@@ -114,7 +114,7 @@ The same arm exists at authoring time: **the workaround — an absorbing fallbac
 - **Model:** DFS the concept DAG before inventing vocabulary; fact-bundle modeling (invent or reuse on proven coincidence, never bare-alias); a finished stage is one fold (any non-fold residue is either a named irreducible kernel or un-migrated modeling — there is no third); model just-in-time and let the mark on the carrier be the authority (no parallel-ledger docs); **no scaffold lands by author declaration alone** — every net-new scaffold is represented by one exact dissolution obligation, joined to an external operator verdict on the current head, and refused in its absence (§5); reviewers independently search for undeclared scaffolds, so the author's classification is evidence to inspect, never the denominator.
 - **Intellectual sustainability:** don't spend future author, reviewer, operator or maintainer time to buy present convenience. Work expected to be thrown away is *presumed redundant* (§2), so back up and complete the construction that will survive; use a temporary one only after the operator explicitly chooses that future cost. The reviewer's independent test, applied whether or not the author labelled anything: **will this artifact survive the terminal architecture substantially unchanged, and be consumed by it?** If no, presume scaffold and stop the merge until the final construction lands or the operator approves the exact exception. A *missing* dissolution condition makes the finding **more** severe, not less — the response is to name the undeclared dissolvable concept, request the final architecture, and require the exact obligation model if the author believes an exception is warranted; adding a trigger after review does not resolve the objection, it only makes the proposal eligible for a decision. Automated detection can never be the whole policy, because temporary work can be authored without using any of the names a gate matches on. Tells, each carrying a presumption: a hand-authored workflow, deployment script, migration script or operational command (out-of-band actuation); a second path beside an existing modeled route (parallel authority); "bridge", "shim", "compatibility", "for now", "temporary", "until", "later" (deferred refactor); a model to be deleted whole when the real one lands (throwaway model); a hand-authored projection the model should generate (manual application committed as source); raw shell implementing semantics already expressible in `.dag` (unmodeled realization); a broad wrapper around a type or modeling deficit (workaround hiding substrate work); a condition whose terminal is "rewrite this properly" (the proper work was not done); a new artifact with no final consumer (experimental residue). If dissolution approvals become frequent, that is not a process cost to optimize away — it is evidence the repository is routinely borrowing future intellectual labor to ship present convenience.
 - **Prioritize holistically, not by the bottleneck:** balance the quantitative and the qualitative — don't anchor on one KPI (you'll hit it at a cost) or on pure taste. Map the cause→effect across sections; a 5ms step doesn't get a pass for not being the 80s one (it might be a 5ns step). One consequence is a standing rule, **bare minimum cost** (operator ruling 2026-07-10): a proven cost-shape defect — a copied accumulator, a quadratic fold — is *always fixed*, regardless of the realized n. "n is small here" is not a time-stable fact (§1/A3 — reuse changes n), and pricing per-site exceptions is itself redundant work (§2); the humility is not trusting your own "negligible here." Root-cause to the language layer and fix related systems *together* — a local subsystem patch is the forked-logic trap. **Denominate the benefit:** the deliverable is a *displaced cost* (§1's time — a pain someone pays to remove); the lens/substrate is the *mechanism* (the moat), not the product. A lens — or any construction wall — is on-dial exactly insofar as it is the cheapest path to such a pain. Priced in elegance instead, the work is self-referential and unbounded (the purity trap — the economic twin of "never" in §5; an extensible substrate's infinite improvability dissolving its own bound).
-- **Enforce with lenses,** not grep — but **construction first** (§5): a lens is *validation* (it concedes the bad state is writable), so make the class unwritable by single authority where you can and reserve the lens for the unstructurable residue. As a residue mechanism it earns its keep: a pure reader over the same `Node` tree, storing nothing, so a new analysis costs zero substrate edits. Beware the tier where the machinery exists but nothing gates on it — coverage by illusion; an inert lens is itself a lie, so an **executable** hygiene check must keep every lens either wired (a discovered fail-closed witness) or deleted — that backstop runs over the corpus and is *not* superseded by the authoring-time construction-justification judgment, which layers on top of it.
+- **Enforce with lenses,** not grep — but **construction first** (§5): a lens is *validation* (it concedes the bad state is writable), so make the class unwritable by single authority where you can and reserve the lens for the unstructurable residue. As a residue mechanism it earns its keep: a pure reader over the same `Node` tree, storing nothing, so a new analysis costs zero substrate edits. Beware the tier where the machinery exists but nothing gates on it — coverage by illusion; an inert lens is itself a lie. **The corpus-wide backstops that used to police this are DELETED (2026-08-11), and the reason is the rule that replaces them:** the inert-lens reach census and the construction-justification census both ran inside floor witness discovery, so *every* discovery run — on every PR, on regen, in every coordinated worker — had to acquire a whole-corpus module graph in order to answer a question about who authored a lens. That is the §6 cost-shape defect at its purest: the unit of computation was the world, the unit of fact was one module's authorship, and the price was paid by every consumer that wanted a witness roster and nothing else. **Who paid it, split by era, because this document must not assert a superseded population as the present one:** before #8140 the walk was unconditional, so ordinary CI, regen, the falsifier cadence and every coordinated worker all paid; #8140 made it demand-directed, so a discovery-free plan such as regen stopped paying; from #8140 to this deletion the two censuses burdened every remaining discovery-bearing execution. The correction is recorded rather than reworded because a change deleting stale supply-side enforcement must not land a fresh stale assertion in the same diff (review on #8141). Neither census had a consumer that justified the acquisition; the inert-lens half additionally reported through two host builtins whose `.dag` surface was a pair of self-recursive stubs (`fn f() { f() }`) reachable only because the interpreter intercepted them. **What is NOT claimed:** this is a real scope narrowing, not a climb. A newly authored lens with no witness, and a lens recording no `construction_justification`, are both writable again and nothing detects either. The obligation survives as review diligence, which is strictly weaker. **Next-rung trigger:** an authorship fact belongs on the module's own declaration, checked at ingestion where the module is parsed anyway — one module's facts from one module's source — rather than reconstructed corpus-wide by a consumer that wanted something else. Until that lands the class sits at *mitigatable*, declared here rather than left to be rediscovered.
 - *e.g.* one catamorphism `fold_node` is reused by all 7 v2 stages; #4699 dissolved `06_translate` 4,912→3,973 lines (`_go` accumulators 35→0); a 6-line `merge_envs` root fix cut reconcile from 81% of the pipeline to 6% (~2× self-compile) — the symptom recurs wherever the root is unfixed (v2 still hand-rolls `ParseTable` because the Realization carrier is staged, not inhabited).
 
 ## 7. Self-hosting (the principles applied to the compiler itself)

dag/gunbc/cli_run_floor_lens_oracle_scaffold.dag

@@ -3,44 +3,6 @@ module gunbc.cli_run_floor_lens_oracle_scaffold
 import std.dissolution { DissolutionCondition, dissolution_description, unbound_dissolution }
 
 
-data cli_run_floor_lens_legacy_graph_oracle_scaffold: Disposition = Scaffold {
-  dissolves_to: SingleAuthority,
-  bind: DeclarationRef {
-    module_path: "v1_compiler.cli_run",
-    decl_name: "floor_lens_graph_legacy",
-    field: WholeDeclaration
-  }
-}
-
-data cli_run_floor_lens_legacy_walk_oracle_scaffold: Disposition = Scaffold {
-  dissolves_to: SingleAuthority,
-  bind: DeclarationRef {
-    module_path: "v1_compiler.cli_run",
-    decl_name: "inert_lens_modules_legacy",
-    field: WholeDeclaration
-  }
-}
-
-data cli_run_floor_lens_oracle_dissolve_trigger: DissolutionCondition = unbound_dissolution(description: "🟡 dissolve-on: floor_lens_graph_legacy + inert_lens_modules_legacy — the 6A repoint's TEST-SIDE equality oracle (the deleted build_floor_lens_import_graph corpus scan and module-name-grain walk, retained only inside cli_run.rs inert_lens_hygiene_tests as the executing equivalence receipt, the same shape as resolve_transitively_bfs_legacy). DELETES WHEN the legacy-oracle class dissolves (the queued resolve_transitively_bfs_legacy triage deletion) OR cli_run.rs Chunk F retires the HAND census entirely (docs/plans/cli-run-reconcile-defork.md)")
-
-data cli_run_floor_lens_census_walk_hand_rust_scaffold: Disposition = Scaffold {
-  dissolves_to: SingleAuthority,
-  bind: DeclarationRef {
-    module_path: "v1_compiler.cli_run",
-    decl_name: "inert_lens_modules",
-    field: WholeDeclaration
-  }
-}
-
-data cli_run_floor_lens_justification_hand_rust_scaffold: Disposition = Scaffold {
-  dissolves_to: SingleAuthority,
-  bind: DeclarationRef {
-    module_path: "v1_compiler.cli_run",
-    decl_name: "lens_justification_census",
-    field: WholeDeclaration
-  }
-}
-
 data cli_run_floor_lens_read_refusal_hand_rust_scaffold: Disposition = Scaffold {
   dissolves_to: SingleAuthority,
   bind: DeclarationRef {
@@ -59,18 +21,12 @@ data cli_run_floor_lens_observation_hand_rust_scaffold: Disposition = Scaffold {
   }
 }
 
-data cli_run_floor_lens_hand_rust_dissolve_trigger: DissolutionCondition = unbound_dissolution(description: "🟡 dissolve-on: inert_lens_modules + lens_justification_census + refuse_on_module_graph_read_refusals + import_resolution_facts_with_observation — the 6A repoint's PRODUCTION census walk, justification census, typed read-refusal arm, and observation projection in HAND-Rust (counted interim seed growth, net +22 production LOC at landing; the host realization of the v2.lens.module_graph selection-tier reach until the census emits from workflow dag). DISSOLVES WHEN cli_run.rs Chunk F lands (docs/plans/cli-run-reconcile-defork.md — the census emits from workflow dag over v2.lens.module_graph and the HAND arms delete) OR ROADMAP 5-dissolve-patches (gunbc.roadmap_authority ticket) retires the HAND census entirely")
-
-data cli_run_floor_lens_hand_rust_loc_delta_net: Int = 22
+data cli_run_floor_lens_hand_rust_dissolve_trigger: DissolutionCondition = unbound_dissolution(description: "🟡 dissolve-on: refuse_on_module_graph_read_refusals + import_resolution_facts_with_observation — the typed read-refusal arm and the observation projection in HAND-Rust (the host realization of the v2.lens.module_graph fact production until it emits from workflow dag). The census walk and justification census this carrier also tracked are DELETED, not dissolved: the supply-side lens enforcement they served made every floor discovery run acquire a whole-corpus world to answer a question about lens authorship, and the answer had no consumer that justified the acquisition. DISSOLVES WHEN cli_run.rs Chunk F lands (docs/plans/cli-run-reconcile-defork.md) OR ROADMAP 5-dissolve-patches (gunbc.roadmap_authority ticket) retires the HAND fact production entirely")
 
 data cli_run_floor_lens_scaffold_shape_red_control: Disposition = Terminal {
   reason: "RED CONTROL for the scaffold-shape cells in dag/test/claim/cli_run_floor_lens_oracle_witness_test.dag: a Terminal disposition, so the canonical reader v2.lens.disposition_redundancy.disposition_is_terminal must answer true for it while answering false for every Scaffold row above. Without this row the shape assertions could pass on a reader that answered false unconditionally; with it, the reader is proven to discriminate. Not a disposition ABOUT any declaration — it is test data declared beside the rows it controls."
 }
 
 data cli_run_floor_lens_oracle_receipt_plan_anchor: String = "docs/plans/cli-run-reconcile-defork.md"
 
-data cli_run_floor_lens_oracle_equality_receipt_test: String = "facts_walk_matches_legacy_floor_lens_graph_on_live_corpus"
-
-data cli_run_floor_lens_oracle_build_count_receipt_test: String = "lens_census_single_facts_build_receipt"
-
-data cli_run_floor_lens_oracle_read_refusal_receipt_test: String = "unreadable_lens_is_a_read_refusal_not_an_absence"
+data cli_run_floor_lens_oracle_read_refusal_receipt_test: String = "unreadable_source_is_a_read_refusal_not_an_absence"