### Description From the 2026-07-05 performance & architecture review (wasm/per-request fixed-cost section + client item 6). Three independent "small each" wins on per-request fixed cost and deploy size: 1. **Tune the release profile.** The shipped artifact is 53.5 MiB of which ~45.6 MiB is DWARF (measured); code section is 5.3 MiB. Add `lto = "thin"`, `codegen-units = 1`, strip debuginfo from the shipped artifact, add `wasm-opt` to the Fastly build. Expected 15-30% code-size cut plus cross-crate inlining gains (the per-request init below is all cross-crate). 2. **Trim the per-request settings pipeline.** The full pipeline runs per request: config-store chunk reads each SHA-256'd, envelope hash, JSON parse, validator walk, and compilation of every handler regex — all re-validating a blob the CLI already validated at push time. Ship a "trusted blob" version flag that skips runtime validation, drop per-chunk hashing (envelope hash covers it), and dedupe the double-loads (JA4 probe path, snapshot fallbacks, `PartnerRegistry` built twice, geo looked up twice). Also cache the parsed Ed25519 signing key per instance. 3. **Build-time tsjs hashes.** Bundles carry a full content hash in the URL but serve with `max-age=300`, and the guest re-concatenates and re-hashes ~350 KB per HTML page view to build the URL. Serve `?v=`-matching requests with `immutable, max-age=31536000`, and compute all hashes in `build.rs` as constants. ### Effort / expected effect - **Effort:** Small each - **Expected effect:** Per-request fixed cost and deploy size. _Sequencing item 9 of 10._