[kalvinnchau]

The sidebar, profile button and sidebar startup no longer re-render on every message, live head update and preview.

Each subscribed to a whole snapshot that the store replaces on unrelated changes:

• ScopedSidebar read the Solo navigation snapshot for its title and the channel list for the conversation's destination. Solo regions are rebuilt on every message, and the list on every preview.
• ProfileButton read the live snapshot for its roster and status, and useSidebarStartup read it for the roster state. Live heads change on every message and prefetch.

Each now selects only the fact it reads: the Solo title, the conversation's destination, the roster state and the live status. All of these are strings or stable references, so useSyncExternalStore skips the render when they are unchanged. The profile refresh effect keeps the same dependency values.

Blox workstations, main 561265d against this change: 5 Linux and 5 macOS, production builds, Playwright WebKit 26.6 and Chromium 153, 100 channels with 10 tabs, 200 history rows and 100 thread replies. Each machine ran every cell once per variant, interleaved with three other candidate changes.

Medians of 5 runs over a 20 s idle window with traffic (20 messages, 79 typing/liveness events):

	Linux before	Linux after	Linux Δ	macOS before	macOS after	macOS Δ
WebKit, background traffic: frame overrun (ms)	669	625	-7%	1381	1318	-5%
WebKit, background traffic: frames over 50 ms	1	0	-1	1	2	+1
WebKit, background traffic: p99 frame (ms)	37	35	-5%	26	25	-4%
WebKit, open-channel traffic: frame overrun (ms)	301	291	-3%	1442	1392	-3%
WebKit, open-channel traffic: frames over 50 ms	0	1	+1	2	2	+0
WebKit, open-channel traffic: p99 frame (ms)	26	25	-4%	25	26	+4%
Chromium, background traffic: script (ms)	180	164	-9%	324	309	-4%
Chromium, background traffic: task (ms)	931	900	-3%	2247	2270	+1%
Chromium, open-channel traffic: script (ms)	187	172	-8%	318	304	-4%
Chromium, open-channel traffic: task (ms)	746	709	-5%	1896	1873	-1%

Repeated switch latency over 12 switches per run, 5 runs (55 samples):

	Linux before	Linux after	Linux Δ	macOS before	macOS after	macOS Δ
WebKit, background traffic: switch p50 (ms)	434	435	+0%	129	128	-1%
WebKit, open-channel traffic: switch p50 (ms)	521	515	-1%	130	130	+0%
Chromium, background traffic: switch p50 (ms)	125	125	-0%	77	75	-3%
Chromium, background traffic: script per switch (ms)	106	103	-3%	62	61	-3%
Chromium, open-channel traffic: switch p50 (ms)	126	125	-1%	75	77	+2%
Chromium, open-channel traffic: script per switch (ms)	109	106	-3%	62	62	-0%

Switch latency is unchanged (p50 within 3%), and keystroke, sidebar hover and Tab focus to frame stay within 2 ms.

In a later round with the same setup, this change stacked on #1967 lowered Chromium idle script by 8% on Linux (176 → 162 ms) and 5% on macOS (332 → 314 ms), and script per switch by 5% and 6%.

Validation:

• useSidebarStartup.test.tsx: a new test checks that a live heads change does not render the hook again, while a roster change does. It fails on main and passes with this change.
• pnpm check, and Vitest for src/app/shell and src/bundled/channels (1128/1128). Full Vitest: 13,941 passed; browser-host/project-git.test.mjs failed once under full-suite load and passes alone.
• Playwright in Chromium and WebKit: navigation-sidebar, sidebar-reveal, solo-navigation, solo-rename, account-profile, startup (94 passed).

🤖 Generated with Claude Code

[wesbillman]

🤖 Star Lord’s automated source review via Wes’s account, integrating independent Rocket, Mantis and Groot review.

No actionable findings in the four-file change and its affected call paths. The selected values preserve Solo title/destination updates, profile-refresh inputs, startup readiness and independently subscribed recovery controls. The existing stores supply stable roster references and primitive selector results; the added regression test checks that heads-only updates do not rerender the startup hook.

Reviewed head: 7dc7e67
Base: 561265d

Source-only: no tests, builds or app execution were performed. The PR’s performance measurements and test results are author-reported, not independently reproduced here; native behavior, rendered focus/accessibility and human acceptance remain unverified. Public PR description and commit metadata were inspected; the description contains no attached images. This COMMENT is not an approval or merge authorization.

[kazemek21]

Kyle reviewed this PR and the Nexus review-tracker notes and approves with no comments.

Posted by Codex (gpt-6.1-sol) on behalf of kylez.