[helly25]

Preserve the complete lookup experiments and the decision to reject them, so future optimization work can build on measured results. Removing temporary-key allocations improved isolated kernels but failed ordinary Linux command controls.

AG;DR

• Documentation-only follow-up to merged Render lexical filename fields from borrowed path slices #1044; synchronized head 4e8d2bc60d is based on main d546fcb3d8. No outstanding parent PR dependency. Runtime prototypes remain local experiments and are preserved as source snapshots/patches in the evidence.
• O05 from the canonical performance inventory. Compare baseline with compiled-key callback references, then compare baseline/reference/transparent-map lookup together.
• Both alternatives remove 2,000 temporary-key allocations per 1,000 renders. Clean release kernels improve, but Linux ten-worker multi-field command controls regress 6.40% / 8.18% in confirmation. Preserve small wins and counterexamples; do not claim an intrinsic comparator/ABI cause.
• Record 10,800 oracle-checked fresh commands across both hosts/phases, nine interleaved rounds with fastest seven, three sessions, discarded warmups, complete source/binary identities and exact native timing-helper source.
• Linux uses tmpfs with 1/3/10 CPU affinity; macOS uses warm APFS and matching worker allowances without affinity. Reference is the first published Render lexical filename fields from borrowed path slices #1044 head f7fea20, not a claim about adopting either variant on newer main.
• Allocation counters run separately under ASan; no sanitizer time is used. The baseline intentionally fails the zero-allocation diagnostic assertion; both alternatives pass. Seven targeted correctness suites pass on both hosts.
• Update TODO and performance analysis with the rejection and remaining O13/O15 prepared-binding work; register the report in the release site. Final output strings still allocate; no interner, wrapper, second catalog or relaxed performance threshold is proposed.
• Normal documentation commit hooks pass. Normal push hooks run from the cached worktree with the inherited parent source scope; no enforcement bypass.

After #1044 merged, GitHub retargeted this PR to main. The report-only commit was rebased without replaying the merged parent and pushed immediately through normal hooks. Replacement main-context CI is authoritative.