analysis: regenerate the full-day results from the fixed code

The output-boundary fixes added two statistical invariants, so the analysis
was rerun end to end at commit 40f7baa. Every data file is byte-identical to
the previous run: the same 61,331 per-second rows, the same rankings, the
same 8,850 symbol rows, the same three state series, and the same SHA-256 for
each. Only the manifest moves, recording the new commit, the new timestamp,
and 57 invariant checks instead of 55.

The report follows the manifest, and its diagnostic timing sentence now lists
the three measured full-day runs rather than one.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: IpsitKalra

analysis/results/analysis.json

@@ -1,7 +1,7 @@
 {
   "schema_version": "1",
-  "generated_at_utc": "2026-07-27T01:00:39Z",
-  "git_commit": "0fdbd1005ca3582a2e8a60d97640e3971f43ce14",
+  "generated_at_utc": "2026-07-27T02:11:12Z",
+  "git_commit": "40f7baa20854ff3a76323788e381ed0c19bf19f6",
   "git_dirty": "false",
   "dataset": {
     "basename": "07302019.NASDAQ_ITCH50",
@@ -913,7 +913,7 @@
   ],
   "invariants": {
     "all_passed": true,
-    "checks": 55,
+    "checks": 57,
     "failures": [
     ],
     "reconciliation": {

analysis/results/findings_candidates.json

@@ -4,7 +4,7 @@
     "analysis_schema_version": "1",
     "dataset": "07302019.NASDAQ_ITCH50",
     "trading_date": "2019-07-30",
-    "analysis_generated_at_utc": "2026-07-27T01:00:39Z"
+    "analysis_generated_at_utc": "2026-07-27T02:11:12Z"
   },
   "selection_policy": "candidates are generated mechanically from the committed outputs; exactly three are selected for docs/day6_findings.md on statistical support, distinctiveness, interpretability, and absence of unsupported causal claims (decisions D-068)",
   "candidates": [

docs/day6_findings.md

@@ -21,7 +21,7 @@ observer and requires identical statuses, book state, and invariants
   verified before the run
 - Complete frames analyzed: 282,229,684 (the whole file, clean EOF)
 - Engine: FastBook, two passes (decisions D-049, D-061)
-- Commit that produced the results: `0fdbd1005ca3`, clean working tree
+- Commit that produced the results: `40f7baa20854`, clean working tree
 - The file stays outside Git; nothing committed derives from it beyond
   aggregates
 
@@ -66,7 +66,7 @@ columns are named `time_of_day` rather than implying UTC.
 
 ## Statistical invariants
 
-The analysis refuses to publish anything until 55 statistical invariants
+The analysis refuses to publish anything until 57 statistical invariants
 pass. They cover the frame accounting, the per-group mutation arithmetic, the
 execution split, the per-second and per-symbol reductions against every
 global total, the selection rule, and, for each selected symbol, the
@@ -75,7 +75,7 @@ state durations summing to the observed duration, the spread-duration weights
 summing to the two-sided duration, and the monotonicity of the weighted
 quantiles.
 
-All 55 passed. Two independent paths then re-checked the results
+All 57 passed. Two independent paths then re-checked the results
 (decisions D-067):
 
 1. **Inside the run.** Pass two filled a dense per-millisecond counter array
@@ -522,9 +522,14 @@ python3 scripts/summarize_findings.py analysis/results
 python3 scripts/render_analysis.py analysis/results analysis/charts
 ```
 
-Diagnostic run record, macOS development machine (Apple M2, 8 GB):
-147.975 seconds of wall time for both passes over the whole file, peak
-resident set size 1,132,036,096 bytes, zero swaps.
+Diagnostic run record, macOS development machine (Apple M2, 8 GB). Both
+passes over the whole file take about 150 seconds. The measured runs of this
+command on this machine were 147.975 seconds with a peak resident set size of
+1,132,036,096 bytes and zero swaps for the committed configuration, 154.312
+seconds for the full-resolution rerun that writes every state change, and
+145.992 seconds for the twenty-five symbol locked and crossed check. This
+timing is diagnostic and is not comparable with the Day 5 benchmark numbers,
+which were produced under a controlled protocol.
 
 ## Output inventory