# Cut 3 workstream 5 close review

Mode: **Type 1**. Invoked review round: **1**. Reviewer: independent Codex
`/root/ws5_close_review`, 2026-10-31. Candidate:
`1ce7b50718380212405d50fe4b298f909d8a7a6e`; frozen product `891c23c8`.

**Verdict: PASS.** No ranked findings. Accepted WS5 direction is implemented
and its inputs suit the bounded synthetic scaling/session question. This
accepts preliminary evidence for LDG-3144/3145, not complete-suite readiness,
a product semantic change or an external performance claim.

1. Read AGENTS, tickets and accepted amendments, file plan, native protocol
   and spike protocol. Independently counted 116 assignments, 64 executed
   and 52 later pending. PMwR covers every WS5 shape; vectorbt is the seventh
   quantitative peer. Top-five rotation, authoring, custom callbacks and PIT
   remain deferred. `git diff 9137a722 1ce7b507` is empty for product paths,
   frozen dense sources and pre-C1 evidence. Scope guard includes inst/design
   and passed before this review was written. Relevant unchanged rules:
   contracts.md#contract-execution-entry-points,
   #contract-execution-snapshot-trust and #contract-result-return-panels.
2. Executed R 4.6.1 Rscript `--vanilla suite/check.R` on the final record with
   `--session-scratch=true --gut=true` and the declared scaling-input directory.
   Six actual N10 workflows completed; `diff.csv` is empty. Changed-fill,
   false-equivalence, wrong-record, omitted-peer, insufficient-repeat,
   actual cash/DONE reuse, count/partial-prefix and Python retention guts pass.
   Scratch and gut directories are under
   `C:/Users/maxth/AppData/Local/Temp/`, respectively
   `ledgr-ws5-review1-20261031-2148-scratch/` and
   `ledgr-ws5-review1-20261031-2148-gut/`.
3. Checked all 235 completed processes against their 470 raw cold/warm clocks;
   every actual session book/ID/reset control matches retained evidence.
   N1000 ledgr counts are durable3/ephemeral2/spot-FIFO3; each assigned peer
   has five repeats at each N. Seven operator-unrun rows implement the accepted
   curtailment, not engine/resource failure. Other ordinary cells have five
   repeats; stress has one descriptive observation and no ratio. Inspected
   native 10x7200/20x23400 fills and financial controls; recomputed stress and
   N1000 phase peaks from raw process samples. Fixed bounds, native prices and
   decision-only classifications remain intact; reduced sweeps gain no
   full-output ratio. Affordable transformed inputs and synthetic minute keys
   answer this controlled question, without establishing production capacity.
4. Actual timestamps show alternating fresh PMwR/vectorbt first positions;
   15 old valid PMwR processes remain unpooled. Other separate windows and
   unattributed TEMP losses are disclosed. Original failures, 31 scaling
   observations and three R observations are preserved; their retained source,
   directory, process and clock metadata match. Verified restored 70/106 exact
   pins outside TEMP and stable defaults. Source recount: R1788 including62
   executable Qmd lines +494 future =2282/2300; Python619+714=1333/2000;
   PowerShell642/650; non-R Qmd266/450. Ledger arithmetic retains actual
   31478.848s plus conservative verification charge =32011.418/36000s;
   dense19396.083/21600s. Review Native work is conservatively below150s,
   covered by the900s final-verification precharge; no extra debit is needed.
5. Independently rendered the explicit committed record offline from the
   benchmark directory. All19 displayed tables equal the committed HTML;
   one embedded PNG is readable and no external resource is required.
   Actual counts, parity, clocks, sampled-memory gaps/lower bounds, required
   materialization, cold/warm boundaries and PMwR journal-price distinctions
   are clear. Independent HTML: TEMP/
   `ledgr-ws5-review1-20261031-2148-report.html`. Temporary read-only audit
   scripts use that same prefix; no performance repeats or installs ran.

No implementation or ticket edits, commits, merges or pushes were performed.
Pre-existing benchmark README changes and local artifacts were preserved.