Brummel
0e411f1796
Architect drift review of cycle 0031 (node-instance naming; spec5818497, pland890090, implffed8cc). Regression gate: no scripts configured (commands.regression is empty) — the architect is the gate. What holds: the ParamAlias retirement is complete and surgical — ParamAlias, aliases_on, check_alias_indices, leaf_has_unaliased_param, Composite.params, and the Composite::new 5th arg are gone from all code with no orphaned references (only historical specs/plans 0016-0024 mention them, correctly untouched); lib.rs re-exports cleaned. C23 preserved in fact: node_name() flows only into signature_of (fan-in) and collect_params (param path), never into the compilat — it stays name-free, wired by raw index. C11 value-empty holds: a node name is construction identity, injected nowhere into the value vector. param_space() is uniformly <node>.<param>; all call sites migrated in lockstep. 198 tests green, clippy clean. Drift fixed (this commit): the C9 fan-in refinement prose still asserted "the graph view renders each fan-in input as the shortest sibling-unique prefix of its source signature" — that render died with ascii-dag in cycle 0026; signature_of now has zero render consumers. Corrected to state signature_of is construction-phase-only with no render consumer, and that surfacing node identity in the WASM viewer is tracked follow-up. Drift routed to backlog: the node instance name does not reach the graph model (model_to_json emits the type label, so the viewer shows identical [SMA] boxes for named fan-in legs) — a deliberate 0031 scope decision, filed as #58 (idea, relates to #13/#37), not a regression. Cycle 0031 drift-clean after the one fix.