[teamleaderleo]

Leo, 2026-10-28: "There's just loads of nonsense in my sidebar that I'm not even looking at. So how could rendering be this f***ing slow?"

Leo-ask: ask-268456b5a0

What was slow

With the session tree on, a session folded under its parent shows up only in the parent's roll-up (blocked / failed / running / unread counts). But SidebarListView.patchRows treated every patch to a hidden session as a height change. So each agent step, title or preview update (several a second per working agent) ran a full reload: it rebuilt the workspace/group/section maps and laid out every row. The more agents a person has folded away, the more often the sidebar re-lays out rows they can't see.

Fix

SidebarSessionTreeState.patchChangesRollUp(from:to:) asks for a layout only when the hidden session's counted state (SidebarAgentState.of, the same value the roll-up counts) changes. Other patches to a hidden session cost the dictionary write and nothing else. Rows that are on screen patch in place as before.

The PR adds no timers, observers or animation loops.

Test

SidebarSessionTreeTests.aFoldedSessionsStepLaysNothingOutButItsStateDoes: a step or title change on a folded session asks for no layout. A state change does, and so does a session the list hasn't seen before. A session that isn't hidden never does, because its own row patches. Under a lane state, the row's own activity is ignored.

Performance

metric	before	after	delta	harness
full list reloads per hidden-session patch that keeps its state	1	0	-1	code path (patchRowsNow → showModelChange → reload); unit test above

Risk: small, one component. A folded parent's chips now update only when a descendant's agent state changes. That matches what the chips count.

Agent session: agent session

🤖 Generated with Claude Code

---

Summary by cubic

Fixes the sidebar so a session folded under its parent no longer triggers a full re-layout on every step, title, or preview update. Those updates happen several times a second per working agent, so the more agents were folded away, the more often the sidebar re-laid out rows that weren't visible. Now only a change to the hidden session's counted state (the value the parent's roll-up chips display) causes a layout.

• Added SidebarSessionTreeState.patchChangesRollUp(from:to:) to decide whether a patch changes a folded ancestor's roll-up; other patches cost only the dictionary write.
• Rows on screen and sessions not hidden under a fold still patch their own row in place as before.

^{Written for commit c0f5a4e. Summary will update on new commits.}

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review c0f5a4e: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr31783 -f ref=lane/sidebar-offscreen-work -f platform=macos -f backend_mode=local

red: cmux-next checks (god files, concurrency, crash safety, l10n); base red: the same failure on the base a252e92; does not block

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4138963169701 (ad01dee1d0)
• cmux NEXT nightly 1.0.0-nightly.4139084486201 (b5060b494d)
• cmux NEXT nightly 1.0.0-nightly.4139171596001 (fb7e28b51d)
• cmux NEXT nightly 1.0.0-nightly.4139290103001 (a92c6ce841)

[teamleaderleo]

Review (subagent claude-opus-5-5 medium): SHIP at c0f5a4e
Follow-ups: none required. A folded parent's chips count only its sessions' agent state (unread, activity, agentWorking). A step, title or preview change can't move those counts. The previous value comes from the list's own workspace map, which reload fills from every node, hidden sessions included.

[blacksmith-sh]

Found 12 test failures on Blacksmith runners:

Failures

Test	View Logs
/tmp/tmp34ls72za/summary-fail.json (live):/summary-fail.json (live):	View Logs
/tmp/tmpqyx6oo1l/summary-missing.json (live):/summary-missing.json (live):	View Logs
[new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows/ [new_workspaces] debug.new_workspace timed 0 of 50 new workspace rows	View Logs
[new_workspaces] the window draws no frames/[new_workspaces] the window draws no frames	View Logs
[tabs] debug.tab_switch timed 0 of 40 switches/ [tabs] debug.tab_switch timed 0 of 40 switches	View Logs
[tabs] the window draws no frames/[tabs] the window draws no frames	View Logs
main.DeveloperIdAppTests.test_download_prints_the_active_profile_only/ test_download_prints_the_active_profile_only	View Logs
stubs: /System/Library/Frameworks/Security.framework/Versions/A/Security has no stub/ Security has no stub	View Logs
stubs: CoreFoundation.tbd: install-name /System/Library/Frameworks/ CoreFoundation.framework/Versions/A/CoreFoundation, missing 1: ['_CFRetain']/ CoreFoundation, missing 1: ['_CFRetain']	View Logs
stubs: CoreFoundation.tbd: install-name /System/Library/Frameworks/ CoreFoundation.framework/Versions/A/CoreFoundation, missing 1: ['_log2']/ CoreFoundation, missing 1: ['_log2']	View Logs

...and 2 more test failures. View all on Blacksmith

^{Need help on this PR? Tag @codesmith-bot with what you need.}