Defer merged icon redraws during menu tracking

[kiranmagic7]

Summary

Defer merged status-icon redraws while the merged dropdown is actively tracking, then apply one latest redraw after the dropdown closes. This reduces one source of in-tracking UI churn called out in #1399 without changing menu content, provider switching, or icon rendering semantics while the menu is closed.

Context

#1399 and the follow-up bisection around #1408 split the merged-menu freeze evidence into two drivers:

1. The switcher event monitor caused the felt input starvation and is addressed by Gate switcher event monitor pump on hardware event counters #1408.
2. WindowServer datagram-buffer overflow persisted with stock churn and disappeared only when all in-tracking UI churn was suppressed.

This PR targets a narrow part of the second bucket: status-item icon redraws requested while the merged menu is open. Those redraws are not user-visible until the menu closes, so the work can be coalesced safely.

Change

• Skip merged status-icon rendering while shouldMergeIcons and isMergedMenuOpen are both true.
• Remember that a merged icon render was deferred.
• Apply one latest icon render after the merged menu closes.
• Keep animation/blink lifecycle reconciliation active while redraws are deferred.
• Preserve quota-warning onset and expiration timing with one event-driven bypass per transition.
• Reconcile the current logical provider when a warning event arrives during a provider/display mismatch.
• Clear the deferred flag during shutdown teardown.
• Add focused regression coverage for deferral, close flushing, animation timeout, warning timing, and provider mismatch.
• Add a changelog entry under 0.32.6 Unreleased.

Validation

• swift test --filter StatusItemAnimationSignatureTests passes: 6 tests.
• Full swift test passes: 3528 tests in 402 suites.
• make check passes: SwiftFormat clean, SwiftLint found 0 violations in 1038 files.
• git diff --check passes.
• Final autoreview is clean with no accepted/actionable findings.
• Ad-hoc packaged bundle reports exact head 841c92d1.
• Peekaboo: 12 merged-menu open cycles completed in 0.09-0.21s; 20 sustained provider-switch input pairs completed with a 0.101s maximum.

Scope notes

This is intentionally not presented as a full fix for #1399. It complements #1408 by reducing one remaining in-tracking churn source that aligns with the WindowServer-overflow evidence.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ca0e327666

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Run animation-state updates when deferring merged renders

When a store/icon change arrives while the merged menu is open, this new early return makes applyIcon report a skipped render, and updateIcons() immediately returns on that value before reaching its existing updateAnimationState()/updateBlinkingState() calls. In the inspected updateIcons() path, that means a refresh completion during menu tracking no longer stops the loading display-link driver (or a refresh start no longer starts it); after the menu closes the deferred icon is applied, but the stale animation driver can keep ticking/redrawing until the 30s timeout or another icon update happens. The render work can be deferred, but the animation/blink state still needs to be updated for changes observed while tracking.

Useful? React with 👍 / 👎.

[steipete]

Landed as b7772b2.

Proof:

• Focused merged-icon redraw tests: 6 passed.
• Full local suite: 3,528 tests in 402 suites passed.
• make check: passed with zero lint violations.
• Auto Review: clean; no accepted/actionable findings.
• Fresh packaged app: 12 menu-open cycles measured 0.09-0.21s; 20 provider-switch pairs measured at or below 0.101s with Peekaboo.
• Exact head 841c92d: macOS CI, Linux x64, Linux arm64, and GitGuardian all passed.

Caveat: this is the bounded redraw mitigation for #1399; the broader WindowServer report remains open for further evidence.