Under calc_on_order_fills on historical bars, the engine capped each
chart bar at four broker fill events along the 4-tick proximity path.
TV has no such budget: a clean-room probe (committed with its export)
is reproduced trade-for-trade, 2325/2325, by eligibility semantics —
orders placed by a MID-BAR recalc may fill only at the remaining
extreme waypoints (gap-filling at the waypoint price, never at exact
intra-segment levels, never at the close tick), pending cascade market
orders roll to the next bar's open, and orders resting at bar start or
placed at the bar-open recalc keep standard semantics. The budget
produced the right two-cycle cap by accident but the wrong reach: it
exact-level-filled cascade brackets on the final segment (the probe's
+302 over-threaded windows) and truncated busy-bar resting-order fills
TV allows.

Mid-bar provenance is a new pending-order flag set from a scheduler
bar-open marker; the cascade gate is scoped to the historical 4-tick
path (the bar-magnifier path keeps its own semantics); a non-semantic
runaway backstop replaces the budget constant. Four legacy assertions
that encoded the budget behavior are corrected to the TV pin.

Gates: the coof cohort is byte-identical (md5-verified 5148- and
524-trade exemplars; all coof corpus probes clean); probe price
exactness 72.6% -> 100% with exit and pnl p90 collapsing to zero;
corpus 240/11/1 zero movement; full private sweep 0 up / 0 down. A
cascade exit-limit mid-segment asymmetry remains as a documented
residual pending its own discriminating probe.

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