Regenerate docs/design-ledgers.md: pristine-main drift measurement (one file, +15/-1)

[briansrls]

What

Regenerates the committed generated artifacts from pristine main (347413337d) by running the committed producer — gunbc run --source-root dag --source-root src/v2 --entry dag/gunbc/instruments/generated_artifact_gate.dag --function main_wet — with a gunbc binary built from the same tree (the system binary at 00:36Z predated two src/v1 changes on main, so it was rebuilt first). Run executed locally under ctrl-build caps (~3.5 GB RSS observed; exceeds the BuildBuddy VM budget, per the brief).

Drift extent (the measurement)

Of the 17 committed generated artifacts, exactly one drifted: docs/design-ledgers.md, +15/-1, entirely within the "Self-host emission board measurement — declared 2026-08-24" rung-drop row:

• the row body is stale against the current authored text in gunbc.rung_drop (the frontier-citing revision: "THE ROW STAYS AS HISTORICAL EVIDENCE…", the gunbc.self_host_compile_phase_frontier persistence paragraph);
• the projected subsection "Whole-corpus compile-phase frontier receipts — latest validated row" (from gunbc.design_ledgers compile_phase_frontier_blocks, reading gunbc.self_host_compile_phase_frontier self_host_compile_phase_frontier_receipts) was missing entirely.

Every other artifact — DESIGN.md, ROADMAP.md, workflows, hooks, stage0 layout files, plan docs — regenerated byte-identical. main_wet exit 0, working tree diff is exactly this file.

Would the drift gate have caught it?

Yes on content, no on execution — it has no executed route. tools.generated_artifact_gate is composed into gunbc_ci_gates (gunbc.ci_spec) and reads committed bytes against regenerated bytes, so this drift is squarely in its refusal set. But:

• .github/workflows/witnesses.yml never invokes tools.ci_gates; the required run's phases carry no generated-artifact drift phase since the 2026-08-29 cut;
• the required floor executes only discovered test fns — tools.floor_effect_gate_witness generated_artifact_drift_gate_passes is a func with zero test-fn consumers;
• the required regen phase (v2.workflow.required_regen) adjudicates src/v2 Rust mirrors, not the doc/workflow artifact roster;
• the hooks check formatting only.

This absence is exactly the population of the "Required gate reduced to the compiler floor — declared 2026-08-29" §4b(3) drop, whose row names generated-artifact drift among the dropped phases (and tools.ci_gates' own comment block records the earlier 2026-07-26 re-homing → cadence-deletion path that orphaned the gate). Per §4b(3) the drop is retired by its trigger and by nothing else — this PR repairs the drifted bytes and measures the extent; it does not (and should not) re-wire the gate.

🤖 Generated with Claude Code

https://claude.ai/code/session_01A4455fVEYvrKXVQ245XHoF

[gunbai-bot]

Closing as duplicate of #9804, which carries the same pristine-main regeneration of docs/design-ledgers.md plus the in-scope incident receipt on the 2026-08-29 rung-drop row (per operator ruling: the ledger must end with one account of this repair — same shape as #9799/#9797). The drift measurement and drift-gate analysis from this PR's description stand and are consistent with #9804. — sent from deep-tern-874