11 — Final Summary

Macro: PROCESS_DISCOVERY_CORRELATION_RUNTIME_INVENTORY_FIX_AND_DOT_EXECUTION_READINESS · 2026-06-04 · EXECUTION_MODE · Status: PARTIAL (all safe branches done live; owner/operator schema + runtime enablement remain).

What was asked

Move from "Process Discovery can infer structural candidates" to "Process Discovery has a runtime/correlation model that can verify actual runs, fixes the producer blind spot, and prepares safe DOT execution readiness."

What was delivered

• A. Correlation/runtime model — complete (doc 02). Header+detail identity model; decision = new observation ledger, reuse event_outbox.correlation_id, emulate job_queue.run_id. Not dot_iu_command_run (wrong grain).
• B. Inventory fix — LIVE (doc 03). v_axis_process_inventory_v2: on-demand producers visible, pair-grouped, no inflation; KG 18/18 complete pairs; 28 producer-bearing rows; v1 untouched.
• C. Runtime observation substrate — apply-ready, rehearsed birth-free (doc 04). Two ledger tables; held for operator (no infra drift).
• D. Event model — DRAFT live (doc 05). 7 generic process.* types, active=false, constraint-valid, birth-free.
• E. DOT execution readiness — complete, no execution (doc 06). Wrapper spec + one-pair checklist + best pilot (KG).
• F. Discovery views v2 — LIVE (doc 07). 8 views; runtime hardened off the last_executed backfill; correlation generalised; verified vs structural separated.
• G. Rescoring — complete (doc 08). dot:kg = runtime_missing; 10-definitions-at-birth reaffirmed; smallest verify path named.
• H. RP/UI impact — packaged (doc 09). 6 routes + components; no Nuxt math; report-only.
• I. Next macro — chosen (doc 10): PROCESS_DISCOVERY_RUNTIME_OBSERVATION_APPLY_AND_DRY_RUN.

Decisive live findings (live wins)

1. last_executed is a backfill (1 distinct ts, usage_count=0, 157 non-underscore DOTs) — v1 scored runtime off it; v2 ignores it.
2. Blind-spot filter pinned exactly (trigger_type excludes on-demand) and fixed live without inflation (pair grouping).
3. Correlation substrate already exists (event_outbox/event_pending.correlation_id, job_queue.run_id) — the gap is DOT-layer-only.
4. dot_iu_command_run is the IU-command layer, not DOT runtime — DOT runtime needs the new ledger.

Live mutation

YES — 8 additive read-only views + 7 inactive draft event types. birth_registry 1,158,148 before == after == post-commit (rehearsed net-zero, committed, MCP read-back PASS). Apply-ready (not committed): observation ledger DDL. KB doc-provenance births expected from publishing this report. 0 process born · 0 canon · 0 approval approved · 0 taxonomy node · 0 IU edit · 0 DOT executed · 0 workflow run · AX-PROCESS still CANDIDATE.

Live numbers (2026-06-04)

births 1,158,148 · DOT_KG 36 (18 producers / 18 verifiers, 18/18 complete pairs, 0 runtime) · inventory v1 60 / v2 46 (28 producer-bearing) · evidence_graph_v2 113 · candidates 17 · verified_candidates 1 (job:cut) · event_type_registry 52 (7 draft process.*, 0 active) · dot_iu_command_run 55/15 · job_queue 13 (6 finished, 1 run group).

Compliance

No forbidden action. All writes additive + reversible + rehearsed net-zero. No "verified" claim without real runtime + correlation (only job:cut qualifies). No KG-only hardcode (views generalise across families + jobs + workflows). Producers no longer hidden.

Exact blocker

None for the chosen next macro's first step (apply the rehearsed observation DDL — operator-gated only). Full dot:kg verification needs: substrate applied + dry-run/real run logged (runtime) + shared process_run_id (correlation) + process.* events activated + KG owner registered (§37) + owner birth admission. No engineering blocker to the discovery work itself.