Close spy-lean-parity and dividend-sim-module on the qualifying parity v2 replay (sim rung ready) - #103

run.py now copies a constant list of deviations from the sealed v2 text
into the receipt: the run precedes the independent review the
preconditions require, the sealed status fields stay unedited, a late
module emission is refused after engine.run() rather than mid-run, the v2
results use receipt-v2.json/verdict-v2.json, the annotated manifest
evidence class, INFO-level captured engine logging, and the operator-declared
--harness-commit argument. Nothing sealed, no tolerance and no input changes.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

…eviewed)

receipt-v2.json is the unedited output of the documented bwrap two-run
replay (rc5, extension sha256 a53dd5a2..., no network namespace, cleared
environment) run from commit 8c4d7e7.
verdict-v2.json is compare.py --bars against the sealed v2 manifest and
unchanged tolerances; --lean-data re-derivation writes a byte-identical
verdict. Entry 304 @ 323.58 and exit -304 @ 291.69 by native OCO gap-open
fills, engine-posted distributions 0.00 and 428.64, native end cash
90734.08, zero ERROR log lines, two runs equal after UUID normalization.
The receipt carries the declared deviations, including that the replay
precedes the independent review the preconditions require. v1 receipt.json
and verdict.json (BLOCKED) and every gate row are unchanged.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

…engine views, replay history)

compare.py evaluates acceptance_criteria.preconditions as checks under the
sealed verdict rule: a retained review of exactly the files that ran,
completed before the run; the sealed manifest, tolerances and frozen inputs at
run time and, with --lean-data only, at comparison time; the engine pin and
isolation. It pins the sealed v2 manifest and preregistration sha256 and
refuses a v2 receipt bound to any other file, judges the OCO structure from
the engine cache's own order records at OrderAccepted and at run end, and
checks that every reported AccountState after the initial one is a module
emission. The verdict reports preregistration_qualifying and the execution
checks separately.

run.py records started_utc, an optional --review-record, the read-only mount
observation, the engine order views, the configured module count labelled as
configuration, the full replay history (replay-history-v2.json) and revised
deviations; the sizing note is a limitation. fixture_strategy.py reads each
accepted order back from the engine cache. Nothing sealed changes; the
published receipt-v2.json and verdict-v2.json are stale until the re-run.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

…ondition (96/96 execution checks pass)

receipt-v2.json is the unedited output of the documented bwrap two-run replay
(rc5, extension sha256 a53dd5a2..., loopback only, cleared environment,
read-only harness and data mounts) run from commit
f079f6c. verdict-v2.json is compare.py
--lean-data against the sealed v2 manifest and unchanged tolerances: every
execution check passes (entry 304 @ 323.58 and exit -304 @ 291.69 by native
OCO gap-open fills read back from the engine cache, engine-posted
distributions 0.00 and 428.64, native end cash 90734.08, zero ERROR lines,
two runs equal after UUID normalization), and nine of ten precondition checks
pass. precondition_review fails because no retained independent review
preceded this run, so under the sealed verdict rule the verdict is FAIL and
preregistration_qualifying is false. This supersedes the unqualified PASS
published at 9d55cb4. replay-history-v2.json appends this run. v1
receipt.json and verdict.json and every gate row are unchanged.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

…dge OCO timing and legs from the engine

compare.py (precondition_review) no longer reports PASS or
preregistration_qualifying while any recorded earlier v2 replay has
reviewed_before_run other than true, unless the gate owner's acceptance of
deviation first_v2_run_preceded_review (deviation-acceptance-v2.json, not
created here) existed at run time, is unchanged at comparison time, has exactly
the six schema keys, names that deviation, binds the five reviewed harness
files that ran and is dated before started_utc. The replay history is
re-hashed at comparison time (append-only growth only), and a run whose
reviewed files equal an earlier replay's fails.

v2_oco_checks replaces the tautological submission-instant check with the
engine clock (submitted_clock_ns) and each leg's OrderSubmitted and
OrderAccepted ts_event at the decision ts (probe_oco_moo_p4: 96/96 accepted
legs at the decision bar), checks the filled leg per fill_rule_by_open from
engine_at_accept, and requires unique order_refs equal to the intents, four
distinct legs, no other client_order_id in order_events and no denial or
rejection anywhere. final_quantity and receipt intents/fills project strictly.

Also: determinism is recomputed from runs[*].normalized_economic_sha256 and
undeclared_differing_fields; under v2 a skipped check is FAIL; MOO attribution
is computed only for a failing price; run.py records python -I, a read-only
interpreter prefix, user/pid namespace indicators and documented environment
values (compare.py checks them; the rest is declared as
isolation_partially_observed) and derives the sizing limitation from the
intents. Every replay from f079f6c is listed as superseded. No replay was run;
receipt-v2.json and verdict-v2.json stay the superseded f079f6c results.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

…ceptance

Review record (schema spy-parity-v2-harness-review/1): harness cc4503f
independently reviewed and accepted with zero unresolved findings, accepted
limitations listed. Deviation acceptance first_v2_run_preceded_review: the
six earlier replays ran before review and are superseded; the first replay
from the reviewed harness, run after this commit with the committed history,
may qualify under unchanged sealed criteria; no retry from cc4503f can.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>

…ity and dividend-sim-module

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>

…ping-status revision

Result review: the closure receipt now carries mappings_closed per the
gate's preregistered schema, and mapping-status-20260923.json records the
three v2 rows as resolved under the v2 manifest's own rule (the sealed v2
and hash-bound v1 manifests are not edited; declared deviation from the
note's v1 wording). Tests tie the receipt and revision to the qualifying
verdict and receipt. Stale gate-note sentences rewritten as history.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>