docs: dissolve stale determinism parallel-ledger; DESIGN reflects landed gate + dropped Phase B

[gunbai-bot]

Follow-up doc hygiene after the determinism mechanism landed (A/C/D: #6551/#6576/#6594). No code changes — the running system is already correct; this only removes/updates stale docs.

Audit finding

The superseded concepts (NonDetSource, DeterminismClass, OperationDeterminism) exist in zero .dag files — the code moved to the emergent DeterminismAxis + map_keys-root model, and the lens's in-code prose is current. The stale material was all doc-side: a DESIGN §6 parallel-ledger that drifted while the code evolved.

Changes

• delete docs/plans/determinism-mechanism-design.md — its body §0–§9 described the superseded model (NonDetSource label enum / DeterminismClass+OperationDeterminism split / emit_determinism_gate as a "permanent CI backstop" — the gate was deleted in Determinism: construction compile-gate in, x2 sampling gate out (Phase D) #6594). Its own §9 dissolution conditions are met; authority is the carrier (dag/std/determinism.dag, src/v2/lens/determinism.dag).
• DESIGN.md open thread (edited dag/gunbc/design_document.dag, regenerated via main_wet): "P1 landed" → landed + enforced by construction; remaining = E (consolidation + DeterminismAxis→Determinism rename) and F (cross-function, gated on SymbolIndex); Phase B dropped — resource Clock/Entropy + kind: Observation are dead (0 consumers, ~1 real corpus occurrence), so grounding would cement a §3 fork for no displaced cost.
• non-fold-irreducible-residue-catalogue.md: remove 2 rows naming deleted symbols (determinism_class_eq / non_det_source_eq).
• mechanism-inventory-red-controls.md: EmitDeterminismGate row → determinism_compile_gate.
• type-env-single-authority-design.md: repoint the inbound link to the deleted doc.

Verification

• Generated-artifact drift gate: green (main_wet regen, DESIGN.md == expected + red-receipt).
• No dangling references to the deleted doc (repo-wide grep: 0).
• design_document.dag resolves (proven by main_wet execution).

🤖 Generated with Claude Code

[gunbai-bot]

The failing ci is a pre-existing whole-tree compile-clean break in main, not this PR.

This PR is docs-only. git diff origin/main..HEAD --stat is 6 files — 4 plan docs, dag/gunbc/design_document.dag (a 2-line li(text:) edit), and the regenerated DESIGN.md — zero host-effect code.

The sole batch-1 error is:

compile-clean: non-exhaustive match: missing variant(s) SetHostnameCas
✗ FAIL [batch 1] dag_compile_clean_gate_passes (returned Bool(false))

Located at dag/gunbc/fleet_converge_cli.dag :: converge_cli_applied_knob_count — a file this PR does not touch (byte-identical to origin/main; verified). Its match intent.effect handles every Srv3*/DurableApprovalGrant* variant but has no SetHostnameCas arm and no wildcard.

Cause: b3b05f01b (Shell→dag P6 Part 1) added the SetHostnameCas HostEffect variant without updating that match. main's own CI masked it (the introducing merges' affected-sets scoped the fleet_converge compile-clean away); this PR surfaces it because it touches a compiled .dag (design_document.dag), so it runs the full whole-tree compile-clean.

Owned and already fixed by #6625 (adds SetHostnameCas { … } => fallback to that function). Verified as the complete fix: the only other files mentioning the sibling Srv3ReceiptEmit without SetHostnameCas (srv3_host_effect_apply.dag, srv3_host_effect_apply_witness_test.dag) are constructor/declaration false-positives — neither contains a HostEffect match — consistent with the log reporting exactly one non-exhaustive site.

I'm not pushing a duplicate fix into a load-bearing effect lane I don't own (a wrong fallback over an effect coproduct is a §5 masking risk). Once #6625 merges and green main is re-merged here, this PR's compile-clean clears; I'll re-verify the full floor before calling it ready. This PR carries 2 clean APPROVEs and is mergeable=MERGEABLE; the only outstanding gate is CI, which is red solely on the inherited main break above.

— sent from deep-koi-605