chore(mqtt): silence periodic stats log for production; add get mqtt.stats CLI

The 30s "MQTT: Memory" line was useful during the outbox/sync-publish
investigation but is spam for production. Gate the periodic logMemoryStatus()
call in the MQTT task loop behind MQTT_MEMORY_DEBUG (a dedicated diagnostics
flag, not enabled by plain MQTT_DEBUG or production builds) and stop the
heltec_v3 variant from force-enabling MQTT_MEMORY_DEBUG on its observer_mqtt
envs (now commented out to match heltec_v4). logMemoryStatus() itself is kept
intact for opt-in debugging.

Expose the same data on demand via a new `get mqtt.stats` CLI command backed by
MQTTBridge::formatMqttStatsReply(): free/max heap, queue depth, outbox total,
and per-slot publish ok/err counts (1-based, matching the msgs: line). Fits the
160-byte reply buffer at 6 slots; returns "(bridge not running)" when down.

src/helpers/CommonCLI_Observer.cpp

@@ -695,6 +695,8 @@ bool CommonCLI::handleObserverGetCmd(uint32_t sender_timestamp, const char* conf
       start = (int)_atoi(start_arg);
     }
     formatMQTTPresetListReply(reply, 160, start);
+  } else if (memcmp(config, "mqtt.stats", 10) == 0) {
+    MQTTBridge::formatMqttStatsReply(reply, 160);
   } else if (memcmp(config, "mqtt.status", 11) == 0) {
     MQTTBridge::formatMqttStatusReply(reply, 160, &_mqtt_prefs);
   } else if (memcmp(config, "mqtt.packets", 12) == 0) {

src/helpers/bridges/MQTTBridge.cpp

@@ -250,6 +250,45 @@ void MQTTBridge::formatMqttStatusReply(char* buf, size_t bufsize, const MQTTPref
   snprintf(buf + pos, bufsize - pos, ", q:%d", q);
 }
 
+// On-demand publish-health + heap snapshot for the `get mqtt.stats` CLI command.
+// Same data as the (MQTT_MEMORY_DEBUG-only) periodic logMemoryStatus() line, but
+// returned as a reply instead of logged. Per-slot "sN=ok/err": ok = cumulative
+// accepted publishes, err = cumulative failures (socket error / network timeout).
+// Outbox should read ~0 (QoS0 publishes synchronously); a rising err isolates a
+// broker whose uplink is dropping writes.
+void MQTTBridge::formatMqttStatsReply(char* buf, size_t bufsize) {
+  if (buf == nullptr || bufsize == 0) return;
+  if (s_mqtt_bridge_instance == nullptr || !s_mqtt_bridge_instance->_initialized) {
+    snprintf(buf, bufsize, "> (bridge not running)");
+    return;
+  }
+  MQTTBridge* b = s_mqtt_bridge_instance;
+
+  int q = 0;
+#ifdef ESP_PLATFORM
+  if (b->_packet_queue_handle != nullptr) {
+    q = (int)uxQueueMessagesWaiting(b->_packet_queue_handle);
+  }
+#else
+  q = b->_queue_count;
+#endif
+
+  size_t outbox_total = 0;
+  for (int i = 0; i < RUNTIME_MQTT_SLOTS; i++) {
+    if (b->_slots[i].client) outbox_total += b->_slots[i].client->getOutboxSize();
+  }
+
+  int pos = snprintf(buf, bufsize, "> Free=%d Max=%d q:%d/%d Outbox=%u |",
+                     (int)ESP.getFreeHeap(), (int)ESP.getMaxAllocHeap(),
+                     q, MAX_QUEUE_SIZE, (unsigned)outbox_total);
+  for (int i = 0; i < RUNTIME_MQTT_SLOTS && pos < (int)bufsize - 1; i++) {
+    if (!b->_slots[i].enabled || !b->_slots[i].client) continue;
+    pos += snprintf(buf + pos, bufsize - pos, " s%d=%lu/%lu", i + 1,
+                    b->_slots[i].client->getPublishOk(),
+                    b->_slots[i].client->getPublishErr());
+  }
+}
+
 uint8_t MQTTBridge::getLastWifiDisconnectReason() { return s_wifi_disconnect_reason; }
 unsigned long MQTTBridge::getLastWifiDisconnectTime() { return s_wifi_disconnect_time; }
 
@@ -899,14 +938,17 @@ void MQTTBridge::mqttTaskLoop() {
 
     unsigned long now = millis();
 
-    // Periodic heap + outbox snapshot (compiles to nothing unless MQTT_DEBUG is set).
-    // Outbox= is the value to watch: it should sit near 0 on a healthy uplink and
-    // plateau at the configured cap (not climb) during a stall.
+    // Periodic heap + publish-health snapshot. Gated behind MQTT_MEMORY_DEBUG (a
+    // dedicated diagnostics flag, NOT enabled on production or plain MQTT_DEBUG builds)
+    // so it stays off by default — the same data is available on demand via the
+    // `get mqtt.stats` CLI command (formatMqttStatsReply / logMemoryStatus()).
+    #ifdef MQTT_MEMORY_DEBUG
     static unsigned long last_mem_log = 0;
     if (now - last_mem_log >= 30000) {
       last_mem_log = now;
       logMemoryStatus();
     }
+    #endif
 
     bool wifi_just_connected = handleWiFiConnection(now);
     if (wifi_just_connected) {

src/helpers/bridges/MQTTBridge.h

@@ -468,6 +468,9 @@ public:
    *  (for LoRa). Returns false if the bridge is not running. */
   bool ntpDiag(char* reply, size_t reply_size, bool verbose);
   static void formatMqttStatusReply(char* buf, size_t bufsize, const MQTTPrefs* obs);
+  /** On-demand publish-health + heap snapshot for `get mqtt.stats` (per-slot ok/err,
+   *  outbox size, free/max heap, queue depth). */
+  static void formatMqttStatsReply(char* buf, size_t bufsize);
   /** True when WiFi is set and at least one MQTT slot can run (preset + custom host if needed). */
   static bool isConfigValid(const MQTTPrefs* obs);
   static void formatSlotDiagReply(char* buf, size_t bufsize, int slot_index);

variants/heltec_v3/platformio.ini

@@ -127,7 +127,8 @@ build_flags =
   -D MAX_MQTT_BROKERS=3
   -D MQTT_MAX_PACKET_SIZE=1024
   -D MQTT_DEBUG=1
-  -D MQTT_MEMORY_DEBUG=1
+; Periodic 30s heap/pub-stats serial log — enable only for debugging (use `get mqtt.stats` on demand instead).
+;  -D MQTT_MEMORY_DEBUG=1
 ; Keep default observer profile less verbose to reduce runtime contention.
 ;  -D MESH_PACKET_LOGGING=1
 ;  -D MESH_DEBUG=1
@@ -486,7 +487,8 @@ build_flags =
   -D MAX_MQTT_BROKERS=3
   -D MQTT_MAX_PACKET_SIZE=1024
   -D MQTT_DEBUG=1
-  -D MQTT_MEMORY_DEBUG=1
+; Periodic 30s heap/pub-stats serial log — enable only for debugging (use `get mqtt.stats` on demand instead).
+;  -D MQTT_MEMORY_DEBUG=1
 ; Keep default observer profile less verbose to reduce runtime contention.
 ;  -D MESH_PACKET_LOGGING=1
 ;  -D MESH_DEBUG=1