Fix three engine defects exposed by live SIP paper trials; first passed 1x trial recorded - #123

Independent review (2 lenses, each finding verified) confirmed 9 findings:
- Safety: a halt-flagged quote with a zero or crossed price is no longer
  dropped; it fails closed, so its halt signal cannot be lost. A quote is
  dropped as untradable only when every other field is well-formed and its
  symbol is subscribed; malformed or unsubscribed ones still fail the
  transport.
- Drops are counted per reason and per symbol in transport health and
  written to the run output (dropped_quotes); tests show a dropped quote
  leaves the stored quote and its age unchanged and that quote_stale still
  fires when only invalid quotes arrive.
- Evidence wording: the recorded stop cause is callback_failure at
  quote_normalization; a crossed quote is the most likely cause, backed by
  the committed measurements; the unretained 9/56,855 sample is labelled an
  operator observation; the 10:11 explanation is hedged. The transport
  isolation check (restart log lines only 16-27 ms after stop()) and the
  frozen 1e2c96a hash list are now committed with their script.

New defect found by the next runs (integration build 796e3d1, 11:42 and 11:5x
ET): transport.mark_reconciled refused unless the whole event queue was
empty, which on SIP it never is, so every run failed at the first 30 s
reconciliation (traceback retained via an unchanged-runner wrapper). It now
refuses only while an order event is queued.

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

…s closed, SIP reconciliation)

Adds #123's second commit to the 1e2c96a-based integration build: halt-flagged,
malformed or unsubscribed invalid quotes fail closed; drops counted per reason
and symbol and written to the run output; mark_reconciled refuses only while
an order event is queued. Not for merge; engine suites pass on this tree.

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

…(paper rung ready)

Trial adaptive-20260923g ran from a read-only git archive of main 843dc8f
(with the three #123 engine fixes): 300.5 s at 1x on SIP, 5 real paper
orders with one cancel-replace and 4 native fills over two 1-share round trips
(INTC, GOOGL), 134 crossed quotes dropped, reconciled flat with cash and
positions matching, realized -0.28 USD; the read-only broker order readback
matches. The receipt's broker_readiness is now passed (previous block kept
under broker_readiness.previous); the gate moves to established,
native_proven; trading_gates.py reports rung_ready.paper true.

Evidence of engine and broker order flow at 1x with 1-share orders, not of
strategy edge. The leverage-ladder rungs keep their own receipts.

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

…(paper rung ready)

Trial adaptive-20260923g ran from a read-only git archive of main 843dc8f
(with the three #123 engine fixes): 300.5 s at 1x on SIP, 5 real paper
orders with one cancel-replace and 4 native fills over two 1-share round trips
(INTC, GOOGL), 134 crossed quotes dropped, reconciled flat with cash and
positions matching, realized -0.28 USD; the read-only broker order readback
matches. The receipt's broker_readiness is now passed (previous block kept
under broker_readiness.previous); the gate moves to established,
native_proven; trading_gates.py reports rung_ready.paper true.

Evidence of engine and broker order flow at 1x with 1-share orders, not of
strategy edge. The leverage-ladder rungs keep their own receipts.

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

…(paper rung ready) (#128)

* Establish adaptive-paper-broker-trial on a main-built 1x paper trial (paper rung ready)

Trial adaptive-20260923g ran from a read-only git archive of main 843dc8f
(with the three #123 engine fixes): 300.5 s at 1x on SIP, 5 real paper
orders with one cancel-replace and 4 native fills over two 1-share round trips
(INTC, GOOGL), 134 crossed quotes dropped, reconciled flat with cash and
positions matching, realized -0.28 USD; the read-only broker order readback
matches. The receipt's broker_readiness is now passed (previous block kept
under broker_readiness.previous); the gate moves to established,
native_proven; trading_gates.py reports rung_ready.paper true.

Evidence of engine and broker order flow at 1x with 1-share orders, not of
strategy edge. The leverage-ladder rungs keep their own receipts.

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

* Record the replaced precondition in the gate note and date its history; fix the README hash wording

Independent review of #128: the numbers and the full-duration run check out
(147 decisions over 299.04 s, largest gap 2.11 s), but the note still stated
the earlier 1e2c96a-on-main precondition and a pending receipt as current.
The note now records the precondition as replaced by those direct checks,
with the comparison that would overturn it, and labels the older text as
history. The README names git archive 843dc8f as the hashed tree.

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

* Reflect the adaptive paper trial in runtime-target and the dashboard checkpoint

GitHub review threads on #128: catalogs/us-equities/runtime-target.json
broker_status and observability/grand-dashboard/state.json (paper lane and
alpaca-paper-operational gate rows, recorded_at) still described the trial
as not started.

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

* Keep the dashboard state values within the public-token alphabet

progress.py accepts only [A-Za-z0-9_. /+()-] up to 160 characters; the
previous values used ':' and ','. CI caught it (tests.test_grand_dashboard).

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

---------

Co-authored-by: seathatflowsinourveins <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>