Add some clarifying implementation comments

src/metric.cpp

@@ -421,6 +421,10 @@ struct Metric {
     ~ThreadInit() {
       // Accumulate thread-local state into global state before cleanup
       std::unique_lock<std::mutex> _{mutex};
+      // NOTE: registration_version increment is REQUIRED here because:
+      // - Cached render plan has pre-resolved pointers to thread-local state
+      // - When threads disappear, these pointers become invalid
+      // - Cache invalidation forces rebuild with updated pointer sets
       ++Metric::registration_version;
       auto thread_id = std::this_thread::get_id();
 
@@ -994,6 +998,7 @@ struct Metric {
         // Sum thread-local values
         for (auto *state_ptr : instruction.aggregate_counter.thread_states) {
           double value;
+          // NOTE: __atomic_load works on plain double (not atomic<double>)
           __atomic_load(&state_ptr->value, &value, __ATOMIC_RELAXED);
           total_value += value;
         }
@@ -1079,6 +1084,12 @@ struct Metric {
 
     for (size_t i = 0; i < static_text.size(); ++i) {
       // Copy static text into caller's arena
+      // NOTE: This copying is REQUIRED for memory safety:
+      // - static_text lives in cached_plan's arena (persistent across renders)
+      // - dynamic_text lives in caller's arena (single render lifetime)
+      // - output must live entirely in caller's arena for consistent lifetime
+      // Without copying, output would have mixed arena ownership causing
+      // use-after-free
       output.push_back(arena_copy_string(static_text[i], arena));
 
       // Add corresponding dynamic text
@@ -1279,9 +1290,15 @@ update_histogram_buckets_simd(std::span<const double> thresholds,
     __m128d cmp_result = _mm_cmp_pd(x_vec, thresholds_vec, _CMP_LE_OQ);
     __m128i cmp_as_int = _mm_castpd_si128(cmp_result);
     __m128i ones = _mm_set1_epi64x(1);
+    // NOTE: _mm_cmp_pd returns 0xFFFFFFFFFFFFFFFF for true, 0x0000000000000000
+    // for false The AND with ones converts this to 1 or 0 per 64-bit lane
+    // (extracts low bit)
     __m128i increments = _mm_and_si128(cmp_as_int, ones);
 
     // Load current counts and add increments
+    // NOTE: Using unaligned load/store (_mm_loadu_si128/_mm_storeu_si128)
+    // because uint64_t arrays from arena allocator are not guaranteed 16-byte
+    // aligned
     __m128i current_counts = _mm_loadu_si128((__m128i *)&counts[i]);
     __m128i updated_counts = _mm_add_epi64(current_counts, increments);