chore: archive v1.2 milestone — Extended Protocol Coverage

35 traffic classes across 9 protocol families shipped. Archives
ROADMAP, REQUIREMENTS, and phase directories to milestones/v1.2-*.
Updates README with new protocol families, sound design table,
and [groups] TOML config documentation.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.planning/RETROSPECTIVE.md

@@ -74,12 +74,49 @@
 - Manual string building beats TOML encoder when you need inline comments/annotations
 - Config extension is smooth when the original Load() was designed with clean boundaries
 
+## Milestone: v1.2 — Extended Protocol Coverage
+
+**Shipped:** 2026-03-27
+**Phases:** 4 | **Plans:** 7 | **Timeline:** 1 day (2026-03-27)
+**LOC:** ~6,000 Go (+1,325 from v1.1) | **Files:** 19 modified
+
+### What Was Built
+- Removed stale constants and future-proofed test bounds (Nyquist-based validation)
+- Rebalanced all ClassFreqConfigs to major-second ladder in family frequency bands (65-3250 Hz) with Group field
+- 21 new TrafficClass constants and 30 port-matching rules across 7 protocol families
+- 28 new classification subtests covering all new port/protocol/transport combinations
+- 21 ClassFreqConfigs entries with family-coherent sound design
+- Group-ordered --print-config with section headers and [groups] TOML reassignment
+
+### What Worked
+- Cleanup-first approach (Phase 8) prevented cascading test failures throughout the milestone
+- Design-before-code (Phase 9 frequency table locked before any protocol code) avoided retroactive rebalancing
+- Strict phase ordering (constants -> rules -> config) kept three-location atomicity tests green
+- Musical interval spacing (major second ratio 1.122) produced perceptually distinct yet family-coherent sounds
+- Oscillator normalization bug caught and fixed during frequency rebalancing (math.Abs for alternating-sign harmonics)
+
+### What Was Inefficient
+- Phase 9 plan 09-01 ROADMAP checkbox wasn't updated after execution — milestone readiness check showed incomplete
+- REQUIREMENTS.md traceability wasn't updated for Phase 9 completions (FREQ-01 through GRP-04 still showed Pending)
+- STATE.md accumulated merge conflict markers from earlier worktree operations, persisted unnoticed
+
+### Patterns Established
+- Group field as string (not enum) on FreqConfig — extensible without code changes
+- Major-second ladder pattern for family-band frequency allocation
+- groupOrder canonical slice for deterministic section ordering in config output
+- rawConfig Groups map for TOML [groups] table parsing
+
+### Key Lessons
+- Musical interval ratios are better than fixed Hz steps for frequency allocation — prevents critical band masking
+- Group-based config output requires building from effective (post-reassignment) groups, not defaults
+- Cleanup phases are high-ROI when they prevent test failures in every subsequent phase
+
 ## Cross-Milestone Trends
 
-| Metric | v1.0 | v1.1 |
-|--------|------|------|
-| Phases | 4 | 3 |
-| Plans | 11 | 6 |
-| Days | 3 | 1 |
-| LOC | 3,254 | 4,675 |
-| Avg plan duration | ~8 min | ~5 min |
+| Metric | v1.0 | v1.1 | v1.2 |
+|--------|------|------|------|
+| Phases | 4 | 3 | 4 |
+| Plans | 11 | 6 | 7 |
+| Days | 3 | 1 | 1 |
+| LOC | 3,254 | 4,675 | ~6,000 |
+| Avg plan duration | ~8 min | ~5 min | ~4 min |