From 724ad66f65244ecc2de3e09b21ef6367c0939e07 Mon Sep 17 00:00:00 2001
From: Brummel <mould@mycelium.de>
Date: Wed, 3 Jun 2026 11:44:14 +0200
Subject: [PATCH] audit: cycle 0001 tidy (clean)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Architect drift review over d566fed..a476049 (core-streaming-substrate):
no drift, no debt. The shipped substrate matches the ledger:

- C7: closed four-kind Copy-POD Scalar, no dyn Any / no heap on the hot path,
  AnyColumn type-erased edge with the kind-check at push and a monomorphic
  bypass; wrong-kind push rejected leaving the column untouched.
- C8: Column pre-sized via with_capacity (Box<[T]>, never realloc), financial
  index 0=newest, zero-copy Window, None on cold/out-of-range read.
- C5/C1: run_count is the per-series push counter (canonical glossary term;
  total_count appears only as explicit RustAst-historical reference), monotonic,
  never moved by a read; no RefCell/Rc/Arc/unsafe, push is &mut and read is &.
- Out-of-scope discipline kept: no Node/Ctx/firing/sim-loop/ingestion/OHLCV
  leaked in; lib.rs exports exactly the seven promised items.

Regression gate: profile commands.regression is empty — no-op; architect is the
sole gate and is green. carry-on.

Forward note (not drift): AnyColumn has write-side as_*_mut accessors but no
symmetric read-side &Column / type-erased read-window — the freshness/read loop
in the Ctx cycle will need that counterpart. Expected next-cycle scope.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>