diff --git a/CLAUDE.md b/CLAUDE.md
index 4c9d4a6e67..2181fdbb39 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -8,7 +8,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 Perry is a native TypeScript compiler written in Rust that compiles TypeScript source code directly to native executables. It uses SWC for TypeScript parsing and LLVM for code generation.
 
-**Current Version:** 0.5.1372
+**Current Version:** 0.5.1373
 
 
 ## TypeScript Parity Status
diff --git a/Cargo.lock b/Cargo.lock
index 0a7dc3ce5c..e34b98f01f 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -5547,7 +5547,7 @@ checksum = "9b4f627cb1b25917193a259e49bdad08f671f8d9708acfd5fe0a8c1455d87220"
 
 [[package]]
 name = "perry"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "base64",
@@ -5607,14 +5607,14 @@ dependencies = [
 
 [[package]]
 name = "perry-api-manifest"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "serde",
 ]
 
 [[package]]
 name = "perry-audio-miniaudio"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "cc",
  "libc",
@@ -5622,7 +5622,7 @@ dependencies = [
 
 [[package]]
 name = "perry-codegen"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "inkwell",
@@ -5639,7 +5639,7 @@ dependencies = [
 
 [[package]]
 name = "perry-codegen-arkts"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-hir",
@@ -5647,7 +5647,7 @@ dependencies = [
 
 [[package]]
 name = "perry-codegen-glance"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-hir",
@@ -5655,7 +5655,7 @@ dependencies = [
 
 [[package]]
 name = "perry-codegen-js"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-dispatch",
@@ -5664,7 +5664,7 @@ dependencies = [
 
 [[package]]
 name = "perry-codegen-swiftui"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-hir",
@@ -5672,7 +5672,7 @@ dependencies = [
 
 [[package]]
 name = "perry-codegen-wasm"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "base64",
@@ -5684,7 +5684,7 @@ dependencies = [
 
 [[package]]
 name = "perry-codegen-wear-tiles"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-hir",
@@ -5692,7 +5692,7 @@ dependencies = [
 
 [[package]]
 name = "perry-container-compose"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "async-trait",
@@ -5721,14 +5721,14 @@ dependencies = [
 
 [[package]]
 name = "perry-container-e2e"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
 ]
 
 [[package]]
 name = "perry-diagnostics"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "serde",
  "serde_json",
@@ -5736,7 +5736,7 @@ dependencies = [
 
 [[package]]
 name = "perry-dispatch"
-version = "0.5.1372"
+version = "0.5.1373"
 
 [[package]]
 name = "perry-doc-fixture-my-bindings"
@@ -5747,7 +5747,7 @@ dependencies = [
 
 [[package]]
 name = "perry-doc-tests"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "clap",
@@ -5762,7 +5762,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-ads"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "block2",
  "objc2",
@@ -5772,7 +5772,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-argon2"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "argon2",
  "perry-ffi",
@@ -5780,7 +5780,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-axios"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "reqwest",
@@ -5789,7 +5789,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-bcrypt"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "bcrypt",
  "perry-ffi",
@@ -5797,7 +5797,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-better-sqlite3"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "rusqlite",
@@ -5805,7 +5805,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-cheerio"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "scraper",
@@ -5813,7 +5813,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-commander"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "perry-runtime",
@@ -5821,7 +5821,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-cron"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "chrono",
  "cron",
@@ -5831,7 +5831,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-dayjs"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "chrono",
  "perry-ffi",
@@ -5839,7 +5839,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-decimal"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "rust_decimal",
@@ -5847,7 +5847,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-dotenv"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "serde_json",
@@ -5855,7 +5855,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-ethers"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "rand 0.10.1",
@@ -5863,7 +5863,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-events"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "perry-runtime",
@@ -5871,14 +5871,14 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-exponential-backoff"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
 ]
 
 [[package]]
 name = "perry-ext-fastify"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "bytes",
  "http-body-util",
@@ -5896,7 +5896,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-fetch"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "bytes",
  "lazy_static",
@@ -5909,7 +5909,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-http"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "bytes",
  "h2",
@@ -5933,7 +5933,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-ioredis"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "lazy_static",
  "perry-ffi",
@@ -5943,7 +5943,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-jsonwebtoken"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "jsonwebtoken",
@@ -5954,7 +5954,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-lru-cache"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "lru",
  "perry-ffi",
@@ -5963,7 +5963,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-moment"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "chrono",
  "perry-ffi",
@@ -5971,7 +5971,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-mongodb"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "bson",
  "futures-util",
@@ -5983,7 +5983,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-mysql2"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "chrono",
  "perry-ffi",
@@ -5993,7 +5993,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-nanoid"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "nanoid",
  "perry-ffi",
@@ -6002,7 +6002,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-net"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "bytes",
  "perry-ffi",
@@ -6015,7 +6015,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-node-forge"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "const-oid 0.9.6",
  "der 0.7.10",
@@ -6034,7 +6034,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-nodemailer"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "lettre",
  "perry-ffi",
@@ -6044,7 +6044,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-pdf"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "printpdf",
@@ -6052,7 +6052,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-pg"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "sqlx",
@@ -6061,7 +6061,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-ratelimit"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "governor",
  "perry-ffi",
@@ -6069,7 +6069,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-sharp"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "fast_image_resize",
  "image",
@@ -6079,14 +6079,14 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-slugify"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
 ]
 
 [[package]]
 name = "perry-ext-streams"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "lazy_static",
  "perry-ffi",
@@ -6095,7 +6095,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-undici"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "perry-runtime",
@@ -6104,7 +6104,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-uuid"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "uuid",
@@ -6112,7 +6112,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-validator"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ffi",
  "regex",
@@ -6122,7 +6122,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-ws"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "futures-util",
  "lazy_static",
@@ -6135,7 +6135,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ext-zlib"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "brotli",
  "flate2",
@@ -6145,7 +6145,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ffi"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "dashmap",
  "once_cell",
@@ -6154,7 +6154,7 @@ dependencies = [
 
 [[package]]
 name = "perry-hir"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-api-manifest",
@@ -6172,7 +6172,7 @@ dependencies = [
 
 [[package]]
 name = "perry-parser"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-diagnostics",
@@ -6184,7 +6184,7 @@ dependencies = [
 
 [[package]]
 name = "perry-runtime"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "base64",
@@ -6226,14 +6226,14 @@ dependencies = [
 
 [[package]]
 name = "perry-runtime-static"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-runtime",
 ]
 
 [[package]]
 name = "perry-stdlib"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "aes 0.8.4",
  "aes 0.9.1",
@@ -6328,14 +6328,14 @@ dependencies = [
 
 [[package]]
 name = "perry-stdlib-static"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-stdlib",
 ]
 
 [[package]]
 name = "perry-transform"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "perry-hir",
@@ -6344,14 +6344,14 @@ dependencies = [
 
 [[package]]
 name = "perry-ui"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ui-model",
 ]
 
 [[package]]
 name = "perry-ui-android"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "itoa",
@@ -6368,7 +6368,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-geisterhand"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "rand 0.10.1",
  "serde",
@@ -6378,7 +6378,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-gtk4"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "cairo-rs 0.22.0",
@@ -6401,7 +6401,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-ios"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "block2",
@@ -6417,7 +6417,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-macos"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "block2",
@@ -6432,7 +6432,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-model"
-version = "0.5.1372"
+version = "0.5.1373"
 
 [[package]]
 name = "perry-ui-test"
@@ -6443,11 +6443,11 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-testkit"
-version = "0.5.1372"
+version = "0.5.1373"
 
 [[package]]
 name = "perry-ui-tvos"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "block2",
@@ -6463,7 +6463,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-visionos"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "block2",
@@ -6479,7 +6479,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-watchos"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "block2",
  "libc",
@@ -6492,7 +6492,7 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-windows"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "base64",
  "libc",
@@ -6509,14 +6509,14 @@ dependencies = [
 
 [[package]]
 name = "perry-ui-windows-winui"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "perry-ui-windows",
 ]
 
 [[package]]
 name = "perry-updater"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "anyhow",
  "base64",
@@ -6532,7 +6532,7 @@ dependencies = [
 
 [[package]]
 name = "perry-wasm-host"
-version = "0.5.1372"
+version = "0.5.1373"
 dependencies = [
  "wasmi",
 ]
diff --git a/Cargo.toml b/Cargo.toml
index c8023fb523..848cd6cdbe 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -315,7 +315,7 @@ codegen-units = 16
 codegen-units = 16
 
 [workspace.package]
-version = "0.5.1372"
+version = "0.5.1373"
 edition = "2021"
 license = "MIT"
 repository = "https://github.com/PerryTS/perry"
diff --git a/changelog.d/7654-plan-close-7478-json-tape-scan.md b/changelog.d/7654-plan-close-7478-json-tape-scan.md
new file mode 100644
index 0000000000..8847ed80cd
--- /dev/null
+++ b/changelog.d/7654-plan-close-7478-json-tape-scan.md
@@ -0,0 +1,64 @@
+**docs(engine-plan): #7478 — the JSON tape's scan penalty is closed, re-measured at v0.5.1370 (PR #7654).**
+
+`docs/engine-plan.md`'s item 2 still described the JSON lazy tape as costing
+**2.3× the direct parser on a full array scan**, with the v0.5.1299 sweep's
+`field_access` inversion (optimized 2984 ms against idiomatic 1350 ms) as the
+open problem. Every step of the issue's roadmap had shipped since — #7483
+(DirectParser float parity, the step-1 blocker), #7499 (reparse-on-materialize),
+#7537 (early batch-flip trigger), #7539 (tape out of the old generation), plus
+#7546 for the wrong-JSON defect the work surfaced — but nobody had re-run the
+decomposition the issue was written around.
+
+Re-measured on the pinned quiet mini, 11 interleaved rounds, 176/176 samples,
+every arm's checksum byte-identical to node 26.5.1:
+
+| phase | tape on: issue → now | tape off: issue → now |
+|---|--:|--:|
+| parse only | 210 → **160 ms** | 1220 → 1196 ms |
+| parse + full scan | 3030 → **1233 ms** | 1287 → 1224 ms |
+| parse + stringify | 254 → **171 ms** | 1756 → 1449 ms |
+| field_access | 2981 → **1721 ms** | — → 1480 ms |
+
+The headline is gone: a full scan was 2.3× the direct parser and is now
+**1.01×** (1233 vs 1224), while `roundtrip` — the path that must not pay for it
+— improved 254 → 171 ms.
+
+**Why it closes rather than continues.** Measuring all four `PERRY_JSON_TAPE` ×
+`PERRY_GEN_GC` combinations instead of the two-switch `idiomatic` arm the
+issue's floor came from, `field_access` decomposes additively: the tape costs
+~200 ms whichever collector runs (241 under gen-GC, 195 under mark-sweep) and
+the collector costs ~560 ms whether or not the tape exists (588 with, 542
+without). The issue had documented these as an *interaction* — scan σ 214.9
+under gen-GC against 8.8 under mark-sweep for the identical tape — and that
+interaction is what #7539 removed. The ~200 ms is the tape *build* (parse-only
+is 160 ms), structural and already named in #7537: the build is purely additive
+whenever the whole tree ends up materialized anyway, and nothing can predict
+scan-shaped access before the parse. The remaining ~560 ms is a
+generational-collector term the tape-off arm carries identically, so it is
+tracked with the GC work rather than as tape policy.
+
+**Method, recorded in the doc.** The "re-measure before scoping" warning goes
+from three instances to four, and gains a second failure mode this ticket
+demonstrates: **a stale floor is as misleading as a stale headline.** #7478's
+acceptance bar was "materially under the 1350 ms `idiomatic` row", but
+`idiomatic` is a measurement rather than a constant and had itself moved to
+938 ms. Working the ticket as written would have chased a headline that four
+merges had already fixed, and then declared failure against a bar that no
+longer existed. When acceptance is "beat arm X", re-measure arm X in the same
+run.
+
+**Verification.** Both reported runs are on a host gated to `rustc == 0` *and*
+1-minute load < 1.5 held across two consecutive checks, and they agree
+independently; a first attempt was discarded when another agent's build started
+mid-run (load 1.86 → 8.57, every σ 50–600). The harness fails rather than
+blanks on a cell that produced no sample or no checksum and asserts its
+expected sample count, so a pass cannot be confused with "it never ran".
+Subject liveness is asserted both ways: the tape is demonstrably engaged
+(tape-on/tape-off differ 7.5× on parse-only and 8.5× on roundtrip), and
+`cargo test -p perry-runtime --lib json_tape` passes 24/24 including the two
+cases that distinguish *which producer ran* via `reparse_materializations()`.
+Binaries built `-p perry -p perry-runtime-static -p perry-stdlib-static` with
+`PERRY_RUNTIME_DIR` pinned, the archive mtime asserted to have moved, and
+`PERRY_NO_AUTO_OPTIMIZE=1` on every compile.
+
+Documentation only — no runtime or codegen change.
diff --git a/docs/engine-plan.md b/docs/engine-plan.md
index bded91b1bc..5956214624 100644
--- a/docs/engine-plan.md
+++ b/docs/engine-plan.md
@@ -308,48 +308,72 @@ each moved by an order of magnitude instead of a few percent:
   delete-safe index walk over the flat entries it is **5 ms**. Perry already won
   lookup before the fix (76 ms vs node 115, bun 103).
 
-**Read this section before picking up any row above.** Three times this campaign
+**Read this section before picking up any row above.** Four times this campaign
 a ticket was worked from a headline number that had already collapsed — #7510
-(33.6% → 11%), `layout_forget_object` (14.5% → 3.0% → 1.7%), and `layout_note_slot`
-(7.5% → 0.03%, correctly closed with **no code at all**). Re-measure the row
-before profiling it.
+(33.6% → 11%), `layout_forget_object` (14.5% → 3.0% → 1.7%), `layout_note_slot`
+(7.5% → 0.03%, correctly closed with **no code at all**), and #7478 (a 2.3x scan
+penalty that was 1.01x by the time anyone re-ran it). Re-measure the row before
+profiling it.
 
-### JSON polyglot legs — the tape is a net negative on scans
+#7478 adds a second failure mode to watch for: **a stale floor is as
+misleading as a stale headline.** Its acceptance bar was "materially under the
+1350 ms `idiomatic` row" — but `idiomatic` is a *measurement*, not a constant,
+and it had moved to 938 ms. A bar quoted as a number silently becomes a bar
+quoted against a different build. When acceptance is "beat arm X", re-measure
+arm X in the same run, never carry its number forward.
 
-Same run. `roundtrip` is the crown jewel and `field_access` is the problem:
+### JSON polyglot legs — the tape's scan penalty is closed (#7478)
+
+`roundtrip` is the crown jewel and `field_access` was the standing problem: at
+the v0.5.1299 sweep the optimized configuration was 2.2x SLOWER than the
+unoptimized one at 3.6x the RSS, with a sigma of 136 against every other row's
+under 5.
 
 | leg | perry optimized | perry idiomatic | bun | node | rust serde_json |
 |---|--:|--:|--:|--:|--:|
 | roundtrip | **192 ms** (82 MB) | 1307 ms | 216 | 379 | 178 |
 | field_access | **2984 ms** (219 MB, sigma 136) | **1350 ms** (61 MB) | 218 | 380 | 183 |
 
-Perry **wins roundtrip** against both JS runtimes and lands within ~8% of Rust
-serde_json. `field_access` was the standing problem: the optimized configuration
-was 2.2x SLOWER than the unoptimized one at 3.6x the RSS, with a sigma of 136
-against every other row's under 5.
-
-**That inversion is now closed** (#7478 → #7537 early batch flip, then #7539
-tape side-allocation), measured on the same host:
+**That inversion is closed** — #7483 (DirectParser float parity), #7499
+(reparse-on-materialize), #7537 (early batch flip), #7539 (tape
+side-allocation), with #7546 fixing the wrong-JSON defect the work surfaced.
+Re-measured at **v0.5.1370** on the pinned quiet mini, 11 interleaved rounds,
+every arm's checksum identical to node 26.5.1:
 
-| `field_access` | median | sigma | peak RSS |
+| phase | tape on, then → now | tape off, then → now |
+|---|--:|--:|
+| parse only | 210 → **160 ms** | 1220 → 1196 ms |
+| parse + full scan | 3030 → **1233 ms** | 1287 → 1224 ms |
+| parse + stringify | 254 → **171 ms** | 1756 → 1449 ms |
+| field_access | 2981 → **1721 ms** | — → 1480 ms |
+
+The issue's headline was that a full scan cost **2.3x the direct parser**
+(3030 vs 1287). It is now **1.01x** (1233 vs 1224): the tape neither wins nor
+loses a scan, which is the correct resting state for a structure whose whole
+value is on the paths that skip materialization. `roundtrip` — the memcpy path
+this must not regress — went the other way, 254 → 171 ms.
+
+**The two remaining terms are now orthogonal, which is the real result.** The
+issue documented an *interaction* (scan sigma 214.9 under gen-GC vs 8.8 under
+mark-sweep for the identical tape). Measuring all four `PERRY_JSON_TAPE` x
+`PERRY_GEN_GC` combinations, `field_access` decomposes additively:
+
+| | tape on | tape off | tape term |
 |---|--:|--:|--:|
-| sweep (v0.5.1299) | 2984 ms | 136 | 219 MB |
-| after #7537 | 2043 ms | 146 | 195 MB |
-| **after #7539** | **1809 ms** | **17.3** | **155 MB** |
-
-Sigma collapses **8.3x** — that was the headline symptom, not the median — and
-the decisive result is that **turning the tape ON is no longer worse than
-turning it OFF**. The tape-off arm still carries sigma 117.2 and 168 MB, so the
-residual variance and footprint belong to the generational collector's behaviour
-on this workload and have nothing left to do with the tape. The GC trace agrees
-to the cycle: 19 cycles / 9 full / 6 `old_gen_bytes` becomes **14 / 5 / 2**,
-which is the `PERRY_JSON_TAPE=0` arm's profile exactly.
+| gen-GC | 1721 ms | 1480 ms | **+241** |
+| mark-sweep | 1133 ms | 938 ms | **+195** |
+| gen-GC term | **+588** | **+542** | |
 
-`roundtrip` — the memcpy path this must not regress — **improved**, 201 → 193 ms,
-with peak old-generation in-use down 39.6 → 14.1 MB.
+The tape costs ~200 ms whichever collector runs, and the collector costs
+~560 ms whether or not the tape exists. The ~200 ms *is* the tape build (parse
+only is 160 ms) and is structural, exactly as #7537 recorded: the build is
+purely additive whenever the whole tree ends up materialized anyway, and
+nothing can predict scan-shaped access before the parse.
 
-Still open: 1809 ms has not reached the 1350 ms idiomatic floor, and the gap is
-now collector behaviour rather than tape design.
+So `field_access`'s remaining ~560 ms is a **generational-collector** term that
+the tape-off arm carries identically. It is not a tape-policy problem and does
+not belong to #7478; it is the same collector behaviour the GC campaign is
+already working, on a workload that happens to reach it through `JSON.parse`.
 
 ### Gates and blockers
 
@@ -379,8 +403,23 @@ now collector behaviour rather than tape design.
    rooting re-reads are material on a spread-heavy workload. If they are, the
    answer is hoisting them (#7487's pooled-alloca precedent), never removing
    them — they close real use-after-frees.
-2. **#7478 — the JSON tape's scan path**, where our optimized build is 2.2x
-   slower than our unoptimized one. The 1350 ms idiomatic row is the floor.
+2. ~~**#7478 — the JSON tape's scan path**~~ — **closed, re-measured at
+   v0.5.1370.** The headline (a full scan costs 2.3x the direct parser) is
+   gone: 3030 → 1233 ms against the tape-off arm's 1224, i.e. **1.01x**, and
+   `roundtrip` improved 254 → 171 ms rather than paying for it. The four-way
+   `PERRY_JSON_TAPE` x `PERRY_GEN_GC` decomposition above is what closes it —
+   the tape term (~200 ms, its own build) and the collector term (~560 ms) are
+   now **additive and independent**, where the issue had documented them as an
+   interaction. The residual on this benchmark is therefore the collector's,
+   carried identically by the tape-off arm, and is tracked with the GC work
+   rather than here.
+
+   Worth recording as method: **this ticket's headline was stale in both
+   directions.** The 2.3x had been fixed by four merges nobody had re-measured
+   together, and its stated floor (the 1350 ms `idiomatic` row) had itself
+   moved to 938 ms — so coding to the ticket would have chased a target that
+   had already moved and declared failure against a bar that no longer existed.
+   Two switches, measured one at a time, was the whole difference.
 3. ~~**`_tlv_get_addr` — thread-local addressing**~~ — **measured out (#7565).**
    Re-measuring first is what decided the design: the 27.0% was real, but the
    ticket's "41 distinct call-graph sites, this is diffuse" was not — **seven