# Fresh profiling transcript

Before: [fa2f4ff](https://git.cafe/pingdotgg/t3code/commit/fa2f4ff2a2bac5ac92e07749f6b83b0003d642a0). After: [12ac6ac](https://git.cafe/pingdotgg/t3code/commit/12ac6ac5cf5b0154b92af29fc82a45f4b1fa164b).

Linux x86_64; Playwright 1.60.0 Chromium 1223; real Vite development apps from separate checkouts and fresh synthetic homes. 1280×800 CSS pixels, DPR 1, no mobile emulation, 16px root font, dark, reduced motion. 201 launched synthetic threads and 201 rendered row DOM nodes per sample; no paid agent turn. Same React fiber instrumentation both builds. P1 stock web defaults; P2 custom unconditional mod+1..9 in addition to ineligible desktop defaults.

Trusted CDP input queues Control down → J down → J up → Control up at 0/2/4/6 ms protocol timestamps. Renderer handling can lag those timestamps; raw handler-observed intervals are included below. The first-frame metric is time from the capture-phase J keydown listener to the first requestAnimationFrame callback, not measured pixel presentation. React actualDuration is the sum of Sidebar fiber actualDuration on performed commits. Sidebar/SortableThreadRow counters skip bailed-out subtrees. Measurement continues at least 650 ms after action completion, covering delayed renders. Do not extrapolate a numerical speedup to production or other machines.

| Fixture/action | Build | Sidebar renders (3 samples) | Row renders | Sidebar actualDuration ms | First RAF callback ms | Commit window ms |
|---|---|---|---|---|---|---|
| P1 open | before | [4, 4, 4] | [804, 804, 804] | [187.7, 175.6, 152.6] | [470.4, 247.8, 224.6] | [1394.4, 975.3, 945.2] |
| P1 close | before | [4, 4, 4] | [804, 804, 804] | [168.4, 156.8, 162.0] | [199.4, 168.9, 175.5] | [935.6, 898.8, 904] |
| P1 open | after | [0, 0, 0] | [0, 0, 0] | [0, 0, 0] | [55.6, 54.4, 48.4] | [729.1, 731.9, 721.5] |
| P1 close | after | [0, 0, 0] | [0, 0, 0] | [0, 0, 0] | [34.8, 29, 33.5] | [725.3, 708.8, 715.5] |
| P2 open | before | [4, 4, 4] | [804, 804, 804] | [175.7, 179.9, 173.0] | [227.6, 229.6, 231.1] | [1003.1, 965.4, 959.8] |
| P2 close | before | [4, 4, 4] | [804, 804, 804] | [211.9, 156.9, 153.7] | [236.4, 158.9, 158.2] | [984.1, 882.9, 877.4] |
| P2 open | after | [0, 0, 0] | [0, 0, 0] | [0, 0, 0] | [64.3, 74.9, 54.3] | [752.3, 764.7, 729.8] |
| P2 close | after | [0, 0, 0] | [0, 0, 0] | [0, 0, 0] | [29.5, 29.1, 34.2] | [716.5, 712.3, 716.1] |

## Input timing audit

| Fixture/action | Build | Protocol hold ms | Handler-observed hold ms |
|---|---|---|---|
| P1 open | before | [6.0, 6.0, 6.1] | [489.8, 249.5, 223.7] |
| P1 close | before | [6.0, 5.9, 6.0] | [198.6, 173.0, 167.2] |
| P1 open | after | [6, 6, 6] | [58.7, 58.7, 48.5] |
| P1 close | after | [6.1, 6, 6] | [37.2, 31.0, 34.7] |
| P2 open | before | [6, 6, 6] | [276.0, 245.1, 234.1] |
| P2 close | before | [6, 6, 6] | [240.3, 159.3, 157.5] |
| P2 open | after | [5.9, 6.0, 6.1] | [72.0, 76.8, 56.2] |
| P2 close | after | [6.0, 6.0, 6.0] | [33.5, 30.8, 36.6] |

The protocol describes a quick physical interval; the old renderer can process key release hundreds of milliseconds later. The performance claim is zero head Sidebar renders while hint visibility stays false, rather than zero renders for deliberate visible hints, route changes or data updates.

## Controls

before P1:
- idle: Sidebar 0, rows 0, observed window 1160.3 ms; actual focus {'tag': 'BODY', 'terminalOwner': None}.
- isolated-mod: Sidebar 2, rows 402, observed window 860.8 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- long-hold: Sidebar 2, rows 402, observed window 2803.2 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- focus-only: Sidebar 2, rows 402, observed window 754.2 ms; actual focus {'tag': 'TEXTAREA', 'terminalOwner': 'drawer'}.
after P1:
- idle: Sidebar 0, rows 0, observed window 1154.6 ms; actual focus {'tag': 'BODY', 'terminalOwner': None}.
- isolated-mod: Sidebar 0, rows 0, observed window 744.5 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- long-hold: Sidebar 0, rows 0, observed window 1818.5 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- focus-only: Sidebar 0, rows 0, observed window 692.9 ms; actual focus {'tag': 'TEXTAREA', 'terminalOwner': 'drawer'}.
before P2:
- idle: Sidebar 0, rows 0, observed window 1158.0 ms; actual focus {'tag': 'BODY', 'terminalOwner': None}.
- isolated-mod: Sidebar 2, rows 402, observed window 859.4 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- long-hold: Sidebar 3, rows 603, observed window 2997.1 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- focus-only: Sidebar 2, rows 402, observed window 807.7 ms; actual focus {'tag': 'TEXTAREA', 'terminalOwner': 'drawer'}.
after P2:
- idle: Sidebar 0, rows 0, observed window 1158.5 ms; actual focus {'tag': 'BODY', 'terminalOwner': None}.
- isolated-mod: Sidebar 0, rows 0, observed window 747.0 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- long-hold: Sidebar 2, rows 402, observed window 1867.8 ms; actual focus {'tag': 'DIV', 'terminalOwner': None}.
- focus-only: Sidebar 0, rows 0, observed window 700.7 ms; actual focus {'tag': 'TEXTAREA', 'terminalOwner': 'drawer'}.

Anchored clock: both builds start client Date at 2027-01-01 12:00 UTC and advance normally at real speed. A strictly frozen Date prevented deadline hooks from reaching the 200 ms boundary; it was discarded. The future anchor makes all synthetic thread ages the same 52-day class and avoids setup-time minute differences. This is an explicit exception to the requested frozen-clock condition.

AI-assisted; verified with the tests and tools listed above.