The sizing nodes reach the run through the readings channel, beside the wired gates, and their result travels with the run under its family. They exclude no configuration and block no run, which is why a reading can state a refusal without stopping anything.
The engine applies no resizing of its own: the P&L series stays in R, and a sizing reading restates it at your declared scale. Making a size act inside the run would take a per-bar size series, and the shipped catalogue carries no node emitting one.
The seam downstream exists and is exercised. The lifecycle simulation accepts a per-day size multiplier and carries the permutation verdict with its result. For a constant size that verdict comes back as a refusal to judge, since a constant modulator has no alignment to permute — which is the reading a constant scheme calls for, because its effect is the level of risk.
The production installer embeds the strategy engine and stops when its source is missing. For sizing experiments, a generic cross-sectional fallback is available as an explicitly labelled demo; it cannot support a deliverable result. Sizing conclusions require an engine-backed run with declared units.
No figure described above is a statement about future performance, and the harness carries no signal to trade and no recommended size. A result concerns the series that was submitted.