docs: update STATE.md after Phase 31-32 revert

Phases 31-32 caused VU meter and volume slider performance regression:
- VU meters animated slowly
- Volume sliders unresponsive when dragged

Reverted to Phase 30 state. Phases 28-30 optimizations remain stable.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

.planning/STATE.md

@@ -9,12 +9,12 @@ See: .planning/PROJECT.md (updated 2026-03-03)
 
 ## Current Position
 
-Phase: 32 of 32 (State Update Optimization)
-Plan: 3 of 4 complete (Loading State Colocation)
-Status: In progress
-Last activity: 2026-03-05 — Completed 32-03-PLAN.md
+Phase: 30 of 30 (Component Memoization)
+Plan: Complete
+Status: Phases 31-32 reverted due to performance regression
+Last activity: 2026-03-06 — Reverted Phases 31-32
 
-Progress: [█████████░] 85%
+Progress: [████████░░] 80%
 
 ## Performance Metrics
 
@@ -25,9 +25,10 @@ Progress: [█████████░] 85%
 
 **v1.7 Performance Optimization (In Progress):**
 - Phases: 5 (phases 28-32)
-- Phases completed: 4 (Phases 28, 29, 30, 31)
-- Requirements: 19 (17 complete)
-- Plans completed: 9
+- Phases completed: 3 (Phases 28, 29, 30)
+- Phases reverted: 2 (Phases 31, 32 caused VU/slider perf regression)
+- Requirements: 19 (15 complete)
+- Plans completed: 7
 
 ## Accumulated Context
 
@@ -52,34 +53,40 @@ Recent decisions (v1.7):
 - React.memo on consumers completes the optimization chain - 29-01
 - Container component memoization completes memoization chain from Phase 29 - 30-01
 - Default shallow comparison sufficient when props are stable from upstream - 30-01
-- Group selectors by usage pattern (core, track, lookup, metadata) - 31-01
-- shallowEqual with composed selectors for object return comparisons - 31-01
-- 1.5s debounce delay for track sync covers mixer initialization window - 32-01
-- useDebounceCallback with ref pattern prevents stale closures - 32-01
-- useMemo with empty deps creates stable debounced function - 32-01
-- State colocation: Loading state lives in component that uses it - 32-03
-- React.memo on self-contained button components prevents parent re-renders - 32-03
+- Group selectors by usage pattern (core, track, lookup, metadata) - 31-01 - REVERTED
+- shallowEqual with composed selectors for object return comparisons - 31-01 - REVERTED
+- 1.5s debounce delay for track sync covers mixer initialization window - 32-01 - REVERTED
+- useDebounceCallback with ref pattern prevents stale closures - 32-01 - REVERTED
+- useMemo with empty deps creates stable debounced function - 32-01 - REVERTED
+- State colocation: Loading state lives in component that uses it - 32-03 - REVERTED
+- React.memo on self-contained button components prevents parent re-renders - 32-03 - REVERTED
+
+**Phase 31-32 Revert (2026-03-06):**
+- Composed selectors with shallowEqual caused VU meters to animate slowly
+- Phase 32 changes caused volume sliders to be unresponsive
+- Root cause: likely blocking main thread during selector evaluation
+- Phases 28-30 optimizations remain effective and stable
 
 ### Investigation Findings (v1.7)
 
 **Root causes of page freezes:**
 1. VU meter updates at 50-70/sec through Redux - FIXED (Phase 28)
 2. MixersContext creates new reference every render - FIXED (Phase 29)
-3. 18+ individual selectors in useMixerHelper - FIXED (Phase 31)
-4. Triple track sync calls on session join - FIXED (Phase 32-01)
-5. Debounce instance recreation on dependency change - FIXED (Phase 32-01)
+3. 18+ individual selectors in useMixerHelper - REVERTED (Phase 31 caused perf regression)
+4. Triple track sync calls on session join - REVERTED (Phase 32 caused perf regression)
+5. Debounce instance recreation on dependency change - REVERTED (Phase 32 caused perf regression)
 6. Missing React.memo on child components - FIXED (Phase 30)
-7. Redundant mixer categorization dispatches - FIXED (Phase 32-02)
-8. Loading state in parent causing unnecessary re-renders - FIXED (Phase 32-03)
+7. Redundant mixer categorization dispatches - REVERTED (Phase 32 caused perf regression)
+8. Loading state in parent causing unnecessary re-renders - REVERTED (Phase 32 caused perf regression)
 
 **Hotspot files:**
 - `useMixerStore.js:178-185` - VU update dispatch - FIXED
 - `useVuHelpers.js:99-104` - setVuStates object creation - FIXED
 - `mixersSlice.js:145-165` - updateMixer creates new objects
-- `JKSessionScreen.js:463-475` - Triple track sync calls - FIXED
-- `useSessionModel.js:605-614` - Debounced trackChanges - FIXED
+- `JKSessionScreen.js:463-475` - Triple track sync calls - REVERTED
+- `useSessionModel.js:605-614` - Debounced trackChanges - REVERTED
 - `MixersContext.js:10` - Unmemoized context value - FIXED
-- `useMixerHelper.js:291-295` - Unconditional category dispatches - FIXED
+- `useMixerHelper.js:291-295` - Unconditional category dispatches - REVERTED
 
 ### Deferred Issues
 
@@ -104,10 +111,11 @@ Recent decisions (v1.7):
 
 ## Session Continuity
 
-Last session: 2026-03-05
-Stopped at: Completed 32-03-PLAN.md (Loading State Colocation)
+Last session: 2026-03-06
+Stopped at: Reverted Phases 31-32 due to VU/slider performance regression
 Resume file: None
 
 **Next steps:**
-1. Execute 32-04-PLAN.md (Final plan in Phase 32)
-2. Plan Phase 33 for verification/performance testing when all optimizations complete
+1. Investigate root cause of Phase 31-32 performance regression
+2. Consider alternative optimization approaches that don't block real-time UI
+3. Or complete v1.7 with Phases 28-30 optimizations only