diff --git a/esphome/components/api/api.proto b/esphome/components/api/api.proto
index 9273cca2d3..0e385c4a17 100644
--- a/esphome/components/api/api.proto
+++ b/esphome/components/api/api.proto
@@ -2310,11 +2310,13 @@ message ZWaveProxyFrame {
 enum ZWaveProxyRequestType {
   ZWAVE_PROXY_REQUEST_TYPE_SUBSCRIBE = 0;
   ZWAVE_PROXY_REQUEST_TYPE_UNSUBSCRIBE = 1;
+  ZWAVE_PROXY_REQUEST_TYPE_HOME_ID_CHANGE = 2;
 }
 message ZWaveProxyRequest {
   option (id) = 129;
-  option (source) = SOURCE_CLIENT;
+  option (source) = SOURCE_BOTH;
   option (ifdef) = "USE_ZWAVE_PROXY";
 
   ZWaveProxyRequestType type = 1;
+  bytes data = 2 [(pointer_to_buffer) = true];
 }
diff --git a/esphome/components/api/api_pb2.cpp b/esphome/components/api/api_pb2.cpp
index 410ba2334e..0140c60e5b 100644
--- a/esphome/components/api/api_pb2.cpp
+++ b/esphome/components/api/api_pb2.cpp
@@ -3112,6 +3112,27 @@ bool ZWaveProxyRequest::decode_varint(uint32_t field_id, ProtoVarInt value) {
   }
   return true;
 }
+bool ZWaveProxyRequest::decode_length(uint32_t field_id, ProtoLengthDelimited value) {
+  switch (field_id) {
+    case 2: {
+      // Use raw data directly to avoid allocation
+      this->data = value.data();
+      this->data_len = value.size();
+      break;
+    }
+    default:
+      return false;
+  }
+  return true;
+}
+void ZWaveProxyRequest::encode(ProtoWriteBuffer buffer) const {
+  buffer.encode_uint32(1, static_cast<uint32_t>(this->type));
+  buffer.encode_bytes(2, this->data, this->data_len);
+}
+void ZWaveProxyRequest::calculate_size(ProtoSize &size) const {
+  size.add_uint32(1, static_cast<uint32_t>(this->type));
+  size.add_length(2, this->data_len);
+}
 #endif
 
 }  // namespace esphome::api
diff --git a/esphome/components/api/api_pb2.h b/esphome/components/api/api_pb2.h
index ee8472b21c..d71ee9777d 100644
--- a/esphome/components/api/api_pb2.h
+++ b/esphome/components/api/api_pb2.h
@@ -280,6 +280,7 @@ enum UpdateCommand : uint32_t {
 enum ZWaveProxyRequestType : uint32_t {
   ZWAVE_PROXY_REQUEST_TYPE_SUBSCRIBE = 0,
   ZWAVE_PROXY_REQUEST_TYPE_UNSUBSCRIBE = 1,
+  ZWAVE_PROXY_REQUEST_TYPE_HOME_ID_CHANGE = 2,
 };
 #endif
 
@@ -2971,16 +2972,21 @@ class ZWaveProxyFrame final : public ProtoDecodableMessage {
 class ZWaveProxyRequest final : public ProtoDecodableMessage {
  public:
   static constexpr uint8_t MESSAGE_TYPE = 129;
-  static constexpr uint8_t ESTIMATED_SIZE = 2;
+  static constexpr uint8_t ESTIMATED_SIZE = 21;
 #ifdef HAS_PROTO_MESSAGE_DUMP
   const char *message_name() const override { return "z_wave_proxy_request"; }
 #endif
   enums::ZWaveProxyRequestType type{};
+  const uint8_t *data{nullptr};
+  uint16_t data_len{0};
+  void encode(ProtoWriteBuffer buffer) const override;
+  void calculate_size(ProtoSize &size) const override;
 #ifdef HAS_PROTO_MESSAGE_DUMP
   void dump_to(std::string &out) const override;
 #endif
 
  protected:
+  bool decode_length(uint32_t field_id, ProtoLengthDelimited value) override;
   bool decode_varint(uint32_t field_id, ProtoVarInt value) override;
 };
 #endif
diff --git a/esphome/components/api/api_pb2_dump.cpp b/esphome/components/api/api_pb2_dump.cpp
index a5494168f9..c5f1d99dd4 100644
--- a/esphome/components/api/api_pb2_dump.cpp
+++ b/esphome/components/api/api_pb2_dump.cpp
@@ -662,6 +662,8 @@ template<> const char *proto_enum_to_string<enums::ZWaveProxyRequestType>(enums:
       return "ZWAVE_PROXY_REQUEST_TYPE_SUBSCRIBE";
     case enums::ZWAVE_PROXY_REQUEST_TYPE_UNSUBSCRIBE:
       return "ZWAVE_PROXY_REQUEST_TYPE_UNSUBSCRIBE";
+    case enums::ZWAVE_PROXY_REQUEST_TYPE_HOME_ID_CHANGE:
+      return "ZWAVE_PROXY_REQUEST_TYPE_HOME_ID_CHANGE";
     default:
       return "UNKNOWN";
   }
@@ -2161,6 +2163,9 @@ void ZWaveProxyFrame::dump_to(std::string &out) const {
 void ZWaveProxyRequest::dump_to(std::string &out) const {
   MessageDumpHelper helper(out, "ZWaveProxyRequest");
   dump_field(out, "type", static_cast<enums::ZWaveProxyRequestType>(this->type));
+  out.append("  data: ");
+  out.append(format_hex_pretty(this->data, this->data_len));
+  out.append("\n");
 }
 #endif
 
diff --git a/esphome/components/api/api_server.cpp b/esphome/components/api/api_server.cpp
index ff94b92b5d..1d5c06092f 100644
--- a/esphome/components/api/api_server.cpp
+++ b/esphome/components/api/api_server.cpp
@@ -356,6 +356,15 @@ void APIServer::on_update(update::UpdateEntity *obj) {
 }
 #endif
 
+#ifdef USE_ZWAVE_PROXY
+void APIServer::on_zwave_proxy_request(const esphome::api::ProtoMessage &msg) {
+  // We could add code to manage a second subscription type, but, since this message type is
+  //  very infrequent and small, we simply send it to all clients
+  for (auto &c : this->clients_)
+    c->send_message(msg, api::ZWaveProxyRequest::MESSAGE_TYPE);
+}
+#endif
+
 #ifdef USE_ALARM_CONTROL_PANEL
 API_DISPATCH_UPDATE(alarm_control_panel::AlarmControlPanel, alarm_control_panel)
 #endif
diff --git a/esphome/components/api/api_server.h b/esphome/components/api/api_server.h
index 354c764825..627870af1d 100644
--- a/esphome/components/api/api_server.h
+++ b/esphome/components/api/api_server.h
@@ -126,6 +126,9 @@ class APIServer : public Component, public Controller {
 #ifdef USE_UPDATE
   void on_update(update::UpdateEntity *obj) override;
 #endif
+#ifdef USE_ZWAVE_PROXY
+  void on_zwave_proxy_request(const esphome::api::ProtoMessage &msg);
+#endif
 
   bool is_connected() const;
 
diff --git a/esphome/components/spi/__init__.py b/esphome/components/spi/__init__.py
index 894c6d1878..d803ee66dc 100644
--- a/esphome/components/spi/__init__.py
+++ b/esphome/components/spi/__init__.py
@@ -276,9 +276,6 @@ def get_spi_interface(index):
         return ["&SPI", "&SPI1"][index]
     if index == 0:
         return "&SPI"
-    # Following code can't apply to C2, H2 or 8266 since they have only one SPI
-    if get_target_variant() in (VARIANT_ESP32S3, VARIANT_ESP32S2):
-        return "new SPIClass(FSPI)"
     return "new SPIClass(HSPI)"
 
 
diff --git a/esphome/components/sx126x/sx126x.cpp b/esphome/components/sx126x/sx126x.cpp
index cae047d168..f5393c478a 100644
--- a/esphome/components/sx126x/sx126x.cpp
+++ b/esphome/components/sx126x/sx126x.cpp
@@ -217,7 +217,7 @@ void SX126x::configure() {
     this->write_opcode_(RADIO_SET_MODULATIONPARAMS, buf, 4);
 
     // set packet params and sync word
-    this->set_packet_params_(this->payload_length_);
+    this->set_packet_params_(this->get_max_packet_size());
     if (this->sync_value_.size() == 2) {
       this->write_register_(REG_LORA_SYNCWORD, this->sync_value_.data(), this->sync_value_.size());
     }
@@ -236,7 +236,7 @@ void SX126x::configure() {
     this->write_opcode_(RADIO_SET_MODULATIONPARAMS, buf, 8);
 
     // set packet params and sync word
-    this->set_packet_params_(this->payload_length_);
+    this->set_packet_params_(this->get_max_packet_size());
     if (!this->sync_value_.empty()) {
       this->write_register_(REG_GFSK_SYNCWORD, this->sync_value_.data(), this->sync_value_.size());
     }
@@ -274,7 +274,7 @@ void SX126x::set_packet_params_(uint8_t payload_length) {
     buf[2] = (this->preamble_detect_ > 0) ? ((this->preamble_detect_ - 1) | 0x04) : 0x00;
     buf[3] = this->sync_value_.size() * 8;
     buf[4] = 0x00;
-    buf[5] = 0x00;
+    buf[5] = (this->payload_length_ > 0) ? 0x00 : 0x01;
     buf[6] = payload_length;
     buf[7] = this->crc_enable_ ? 0x06 : 0x01;
     buf[8] = 0x00;
@@ -314,6 +314,9 @@ SX126xError SX126x::transmit_packet(const std::vector<uint8_t> &packet) {
   buf[0] = 0xFF;
   buf[1] = 0xFF;
   this->write_opcode_(RADIO_CLR_IRQSTATUS, buf, 2);
+  if (this->payload_length_ == 0) {
+    this->set_packet_params_(this->get_max_packet_size());
+  }
   if (this->rx_start_) {
     this->set_mode_rx();
   } else {
diff --git a/esphome/components/usb_host/__init__.py b/esphome/components/usb_host/__init__.py
index 0fe3310127..de734bf425 100644
--- a/esphome/components/usb_host/__init__.py
+++ b/esphome/components/usb_host/__init__.py
@@ -1,5 +1,6 @@
 import esphome.codegen as cg
 from esphome.components.esp32 import (
+    VARIANT_ESP32P4,
     VARIANT_ESP32S2,
     VARIANT_ESP32S3,
     add_idf_sdkconfig_option,
@@ -47,7 +48,7 @@ CONFIG_SCHEMA = cv.All(
         }
     ),
     cv.only_with_esp_idf,
-    only_on_variant(supported=[VARIANT_ESP32S2, VARIANT_ESP32S3]),
+    only_on_variant(supported=[VARIANT_ESP32S2, VARIANT_ESP32S3, VARIANT_ESP32P4]),
 )
 
 
diff --git a/esphome/components/usb_host/usb_host.h b/esphome/components/usb_host/usb_host.h
index d2ff4da068..e1e96cfce3 100644
--- a/esphome/components/usb_host/usb_host.h
+++ b/esphome/components/usb_host/usb_host.h
@@ -1,7 +1,7 @@
 #pragma once
 
 // Should not be needed, but it's required to pass CI clang-tidy checks
-#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
+#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3) || defined(USE_ESP32_VARIANT_ESP32P4)
 #include "esphome/core/component.h"
 #include <vector>
 #include "usb/usb_host.h"
@@ -9,11 +9,31 @@
 #include <freertos/task.h>
 #include "esphome/core/lock_free_queue.h"
 #include "esphome/core/event_pool.h"
-#include <list>
+#include <atomic>
 
 namespace esphome {
 namespace usb_host {
 
+// THREADING MODEL:
+// This component uses a dedicated USB task for event processing to prevent data loss.
+// - USB Task (high priority): Handles USB events, executes transfer callbacks
+// - Main Loop Task: Initiates transfers, processes completion events
+//
+// Thread-safe communication:
+// - Lock-free queues for USB task -> main loop events (SPSC pattern)
+// - Lock-free TransferRequest pool using atomic bitmask (MCSP pattern)
+//
+// TransferRequest pool access pattern:
+// - get_trq_() [allocate]: Called from BOTH USB task and main loop threads
+//   * USB task: via USB UART input callbacks that restart transfers immediately
+//   * Main loop: for output transfers and flow-controlled input restarts
+// - release_trq() [deallocate]: Called from main loop thread only
+//
+// The multi-threaded allocation is intentional for performance:
+// - USB task can immediately restart input transfers without context switching
+// - Main loop controls backpressure by deciding when to restart after consuming data
+// The atomic bitmask ensures thread-safe allocation without mutex blocking.
+
 static const char *const TAG = "usb_host";
 
 // Forward declarations
@@ -98,13 +118,7 @@ class USBClient : public Component {
   friend class USBHost;
 
  public:
-  USBClient(uint16_t vid, uint16_t pid) : vid_(vid), pid_(pid) { init_pool(); }
-
-  void init_pool() {
-    this->trq_pool_.clear();
-    for (size_t i = 0; i != MAX_REQUESTS; i++)
-      this->trq_pool_.push_back(&this->requests_[i]);
-  }
+  USBClient(uint16_t vid, uint16_t pid) : vid_(vid), pid_(pid), trq_in_use_(0) {}
   void setup() override;
   void loop() override;
   // setup must happen after the host bus has been setup
@@ -126,10 +140,13 @@ class USBClient : public Component {
 
  protected:
   bool register_();
-  TransferRequest *get_trq_();
+  TransferRequest *get_trq_();  // Lock-free allocation using atomic bitmask (multi-consumer safe)
   virtual void disconnect();
   virtual void on_connected() {}
-  virtual void on_disconnected() { this->init_pool(); }
+  virtual void on_disconnected() {
+    // Reset all requests to available (all bits to 0)
+    this->trq_in_use_.store(0);
+  }
 
   // USB task management
   static void usb_task_fn(void *arg);
@@ -143,7 +160,11 @@ class USBClient : public Component {
   int state_{USB_CLIENT_INIT};
   uint16_t vid_{};
   uint16_t pid_{};
-  std::list<TransferRequest *> trq_pool_{};
+  // Lock-free pool management using atomic bitmask (no dynamic allocation)
+  // Bit i = 1: requests_[i] is in use, Bit i = 0: requests_[i] is available
+  // Supports multiple concurrent consumers (both threads can allocate)
+  // Single producer for deallocation (main loop only)
+  std::atomic<uint16_t> trq_in_use_;
   TransferRequest requests_[MAX_REQUESTS]{};
 };
 class USBHost : public Component {
diff --git a/esphome/components/usb_host/usb_host_client.cpp b/esphome/components/usb_host/usb_host_client.cpp
index 5c9d56c7f9..1cfe2f5f13 100644
--- a/esphome/components/usb_host/usb_host_client.cpp
+++ b/esphome/components/usb_host/usb_host_client.cpp
@@ -1,5 +1,5 @@
 // Should not be needed, but it's required to pass CI clang-tidy checks
-#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
+#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3) || defined(USE_ESP32_VARIANT_ESP32P4)
 #include "usb_host.h"
 #include "esphome/core/log.h"
 #include "esphome/core/hal.h"
@@ -7,6 +7,7 @@
 
 #include <cinttypes>
 #include <cstring>
+#include <atomic>
 namespace esphome {
 namespace usb_host {
 
@@ -185,9 +186,11 @@ void USBClient::setup() {
     this->mark_failed();
     return;
   }
-  for (auto *trq : this->trq_pool_) {
-    usb_host_transfer_alloc(64, 0, &trq->transfer);
-    trq->client = this;
+  // Pre-allocate USB transfer buffers for all slots at startup
+  // This avoids any dynamic allocation during runtime
+  for (size_t i = 0; i < MAX_REQUESTS; i++) {
+    usb_host_transfer_alloc(64, 0, &this->requests_[i].transfer);
+    this->requests_[i].client = this;  // Set once, never changes
   }
 
   // Create and start USB task
@@ -347,17 +350,39 @@ static void control_callback(const usb_transfer_t *xfer) {
   queue_transfer_cleanup(trq, EVENT_CONTROL_COMPLETE);
 }
 
+// THREAD CONTEXT: Called from both USB task and main loop threads (multi-consumer)
+// - USB task: USB UART input callbacks restart transfers for immediate data reception
+// - Main loop: Output transfers and flow-controlled input restarts after consuming data
+//
+// THREAD SAFETY: Lock-free using atomic compare-and-swap on bitmask
+// This multi-threaded access is intentional for performance - USB task can
+// immediately restart transfers without waiting for main loop scheduling.
 TransferRequest *USBClient::get_trq_() {
-  if (this->trq_pool_.empty()) {
-    ESP_LOGE(TAG, "Too many requests queued");
-    return nullptr;
+  uint16_t mask = this->trq_in_use_.load(std::memory_order_relaxed);
+
+  // Find first available slot (bit = 0) and try to claim it atomically
+  for (size_t i = 0; i < MAX_REQUESTS; i++) {
+    if (!(mask & (1U << i))) {
+      // Slot i appears available, try to claim it atomically
+      uint16_t expected = mask;
+      uint16_t desired = mask | (1U << i);  // Set bit i to mark as in-use
+
+      if (this->trq_in_use_.compare_exchange_weak(expected, desired, std::memory_order_acquire,
+                                                  std::memory_order_relaxed)) {
+        // Successfully claimed slot i - prepare the TransferRequest
+        auto *trq = &this->requests_[i];
+        trq->transfer->context = trq;
+        trq->transfer->device_handle = this->device_handle_;
+        return trq;
+      }
+      // Another thread claimed this slot, retry with updated mask
+      mask = expected;
+      i--;  // Retry the same index with new mask value
+    }
   }
-  auto *trq = this->trq_pool_.front();
-  this->trq_pool_.pop_front();
-  trq->client = this;
-  trq->transfer->context = trq;
-  trq->transfer->device_handle = this->device_handle_;
-  return trq;
+
+  ESP_LOGE(TAG, "Too many requests queued (all %d slots in use)", MAX_REQUESTS);
+  return nullptr;
 }
 void USBClient::disconnect() {
   this->on_disconnected();
@@ -370,6 +395,8 @@ void USBClient::disconnect() {
   this->device_addr_ = -1;
 }
 
+// THREAD CONTEXT: Called from main loop thread only
+// - Used for device configuration and control operations
 bool USBClient::control_transfer(uint8_t type, uint8_t request, uint16_t value, uint16_t index,
                                  const transfer_cb_t &callback, const std::vector<uint8_t> &data) {
   auto *trq = this->get_trq_();
@@ -425,6 +452,9 @@ static void transfer_callback(usb_transfer_t *xfer) {
 }
 /**
  * Performs a transfer input operation.
+ * THREAD CONTEXT: Called from both USB task and main loop threads!
+ * - USB task: USB UART input callbacks call start_input() which calls this
+ * - Main loop: Initial setup and other components
  *
  * @param ep_address The endpoint address.
  * @param callback The callback function to be called when the transfer is complete.
@@ -451,6 +481,9 @@ void USBClient::transfer_in(uint8_t ep_address, const transfer_cb_t &callback, u
 
 /**
  * Performs an output transfer operation.
+ * THREAD CONTEXT: Called from main loop thread only
+ * - USB UART output uses defer() to ensure main loop context
+ * - Modbus and other components call from loop()
  *
  * @param ep_address The endpoint address.
  * @param callback The callback function to be called when the transfer is complete.
@@ -483,7 +516,28 @@ void USBClient::dump_config() {
                 "  Product id %04X",
                 this->vid_, this->pid_);
 }
-void USBClient::release_trq(TransferRequest *trq) { this->trq_pool_.push_back(trq); }
+// THREAD CONTEXT: Only called from main loop thread (single producer for deallocation)
+// - Via event processing when handling EVENT_TRANSFER_COMPLETE/EVENT_CONTROL_COMPLETE
+// - Directly when transfer submission fails
+//
+// THREAD SAFETY: Lock-free using atomic AND to clear bit
+// Single-producer pattern makes this simpler than allocation
+void USBClient::release_trq(TransferRequest *trq) {
+  if (trq == nullptr)
+    return;
+
+  // Calculate index from pointer arithmetic
+  size_t index = trq - this->requests_;
+  if (index >= MAX_REQUESTS) {
+    ESP_LOGE(TAG, "Invalid TransferRequest pointer");
+    return;
+  }
+
+  // Atomically clear bit i to mark slot as available
+  // fetch_and with inverted bitmask clears the bit atomically
+  uint16_t bit = 1U << index;
+  this->trq_in_use_.fetch_and(~bit, std::memory_order_release);
+}
 
 }  // namespace usb_host
 }  // namespace esphome
diff --git a/esphome/components/usb_host/usb_host_component.cpp b/esphome/components/usb_host/usb_host_component.cpp
index 682026a9c5..fb19239c73 100644
--- a/esphome/components/usb_host/usb_host_component.cpp
+++ b/esphome/components/usb_host/usb_host_component.cpp
@@ -1,5 +1,5 @@
 // Should not be needed, but it's required to pass CI clang-tidy checks
-#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
+#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3) || defined(USE_ESP32_VARIANT_ESP32P4)
 #include "usb_host.h"
 #include <cinttypes>
 #include "esphome/core/log.h"
diff --git a/esphome/components/usb_uart/__init__.py b/esphome/components/usb_uart/__init__.py
index 6999b1b955..a852e1f78b 100644
--- a/esphome/components/usb_uart/__init__.py
+++ b/esphome/components/usb_uart/__init__.py
@@ -24,7 +24,6 @@ usb_uart_ns = cg.esphome_ns.namespace("usb_uart")
 USBUartComponent = usb_uart_ns.class_("USBUartComponent", Component)
 USBUartChannel = usb_uart_ns.class_("USBUartChannel", UARTComponent)
 
-
 UARTParityOptions = usb_uart_ns.enum("UARTParityOptions")
 UART_PARITY_OPTIONS = {
     "NONE": UARTParityOptions.UART_CONFIG_PARITY_NONE,
diff --git a/esphome/components/usb_uart/ch34x.cpp b/esphome/components/usb_uart/ch34x.cpp
index 601bfe7366..889366b579 100644
--- a/esphome/components/usb_uart/ch34x.cpp
+++ b/esphome/components/usb_uart/ch34x.cpp
@@ -1,4 +1,4 @@
-#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
+#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3) || defined(USE_ESP32_VARIANT_ESP32P4)
 #include "usb_uart.h"
 #include "usb/usb_host.h"
 #include "esphome/core/log.h"
@@ -72,6 +72,7 @@ void USBUartTypeCH34X::enable_channels() {
     if (channel->index_ >= 2)
       cmd += 0xE;
     this->control_transfer(USB_VENDOR_DEV | usb_host::USB_DIR_OUT, cmd, value, (factor << 8) | divisor, callback);
+    this->control_transfer(USB_VENDOR_DEV | usb_host::USB_DIR_OUT, cmd + 3, 0x80, 0, callback);
   }
   USBUartTypeCdcAcm::enable_channels();
 }
diff --git a/esphome/components/usb_uart/cp210x.cpp b/esphome/components/usb_uart/cp210x.cpp
index 35834c7529..5fec0bed02 100644
--- a/esphome/components/usb_uart/cp210x.cpp
+++ b/esphome/components/usb_uart/cp210x.cpp
@@ -1,4 +1,4 @@
-#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
+#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3) || defined(USE_ESP32_VARIANT_ESP32P4)
 #include "usb_uart.h"
 #include "usb/usb_host.h"
 #include "esphome/core/log.h"
diff --git a/esphome/components/usb_uart/usb_uart.cpp b/esphome/components/usb_uart/usb_uart.cpp
index 8603e28d62..29003e071e 100644
--- a/esphome/components/usb_uart/usb_uart.cpp
+++ b/esphome/components/usb_uart/usb_uart.cpp
@@ -1,5 +1,5 @@
 // Should not be needed, but it's required to pass CI clang-tidy checks
-#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
+#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3) || defined(USE_ESP32_VARIANT_ESP32P4)
 #include "usb_uart.h"
 #include "esphome/core/log.h"
 #include "esphome/components/uart/uart_debugger.h"
@@ -216,9 +216,16 @@ void USBUartComponent::dump_config() {
 void USBUartComponent::start_input(USBUartChannel *channel) {
   if (!channel->initialised_.load() || channel->input_started_.load())
     return;
-  // Note: This function is called from both USB task and main loop, so we cannot
-  // directly check ring buffer space here. Backpressure is handled by the chunk pool:
-  // when exhausted, USB input stops until chunks are freed by the main loop
+  // THREAD CONTEXT: Called from both USB task and main loop threads
+  // - USB task: Immediate restart after successful transfer for continuous data flow
+  // - Main loop: Controlled restart after consuming data (backpressure mechanism)
+  //
+  // This dual-thread access is intentional for performance:
+  // - USB task restarts avoid context switch delays for high-speed data
+  // - Main loop restarts provide flow control when buffers are full
+  //
+  // The underlying transfer_in() uses lock-free atomic allocation from the
+  // TransferRequest pool, making this multi-threaded access safe
   const auto *ep = channel->cdc_dev_.in_ep;
   // CALLBACK CONTEXT: This lambda is executed in USB task via transfer_callback
   auto callback = [this, channel](const usb_host::TransferStatus &status) {
@@ -297,7 +304,8 @@ static void fix_mps(const usb_ep_desc_t *ep) {
   if (ep != nullptr) {
     auto *ep_mutable = const_cast<usb_ep_desc_t *>(ep);
     if (ep->wMaxPacketSize > 64) {
-      ESP_LOGW(TAG, "Corrected MPS of EP %u from %u to 64", ep->bEndpointAddress, ep->wMaxPacketSize);
+      ESP_LOGW(TAG, "Corrected MPS of EP 0x%02X from %u to 64", static_cast<uint8_t>(ep->bEndpointAddress & 0xFF),
+               ep->wMaxPacketSize);
       ep_mutable->wMaxPacketSize = 64;
     }
   }
@@ -314,7 +322,7 @@ void USBUartTypeCdcAcm::on_connected() {
   for (auto *channel : this->channels_) {
     if (i == cdc_devs.size()) {
       ESP_LOGE(TAG, "No configuration found for channel %d", channel->index_);
-      this->status_set_warning(LOG_STR("No configuration found for channel"));
+      this->status_set_warning("No configuration found for channel");
       break;
     }
     channel->cdc_dev_ = cdc_devs[i++];
diff --git a/esphome/components/usb_uart/usb_uart.h b/esphome/components/usb_uart/usb_uart.h
index b41e0a52e9..a5e7905ac5 100644
--- a/esphome/components/usb_uart/usb_uart.h
+++ b/esphome/components/usb_uart/usb_uart.h
@@ -1,6 +1,6 @@
 #pragma once
 
-#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3)
+#if defined(USE_ESP32_VARIANT_ESP32S2) || defined(USE_ESP32_VARIANT_ESP32S3) || defined(USE_ESP32_VARIANT_ESP32P4)
 #include "esphome/core/component.h"
 #include "esphome/core/helpers.h"
 #include "esphome/components/uart/uart_component.h"
diff --git a/esphome/components/zwave_proxy/zwave_proxy.cpp b/esphome/components/zwave_proxy/zwave_proxy.cpp
index feaf6e2d42..70932da87c 100644
--- a/esphome/components/zwave_proxy/zwave_proxy.cpp
+++ b/esphome/components/zwave_proxy/zwave_proxy.cpp
@@ -1,4 +1,5 @@
 #include "zwave_proxy.h"
+#include "esphome/components/api/api_server.h"
 #include "esphome/core/application.h"
 #include "esphome/core/helpers.h"
 #include "esphome/core/log.h"
@@ -97,12 +98,19 @@ void ZWaveProxy::process_uart_() {
       // - buffer_[3]: Command ID (0x20 for GET_NETWORK_IDS)
       if (this->buffer_[3] == ZWAVE_COMMAND_GET_NETWORK_IDS && this->buffer_[2] == ZWAVE_COMMAND_TYPE_RESPONSE &&
           this->buffer_[1] >= ZWAVE_MIN_GET_NETWORK_IDS_LENGTH && this->buffer_[0] == ZWAVE_FRAME_TYPE_START) {
-        // Extract the 4-byte Home ID starting at offset 4
+        // Store the 4-byte Home ID, which starts at offset 4, and notify connected clients if it changed
         // The frame parser has already validated the checksum and ensured all bytes are present
-        std::memcpy(this->home_id_.data(), this->buffer_.data() + 4, this->home_id_.size());
-        this->home_id_ready_ = true;
-        ESP_LOGI(TAG, "Home ID: %s",
-                 format_hex_pretty(this->home_id_.data(), this->home_id_.size(), ':', false).c_str());
+        if (this->set_home_id(&this->buffer_[4])) {
+          api::ZWaveProxyRequest msg;
+          msg.type = api::enums::ZWAVE_PROXY_REQUEST_TYPE_HOME_ID_CHANGE;
+          msg.data = this->home_id_.data();
+          msg.data_len = this->home_id_.size();
+          if (api::global_api_server != nullptr) {
+            // We could add code to manage a second subscription type, but, since this message is
+            //  very infrequent and small, we simply send it to all clients
+            api::global_api_server->on_zwave_proxy_request(msg);
+          }
+        }
       }
       ESP_LOGV(TAG, "Sending to client: %s", YESNO(this->api_connection_ != nullptr));
       if (this->api_connection_ != nullptr) {
@@ -120,7 +128,12 @@ void ZWaveProxy::process_uart_() {
   }
 }
 
-void ZWaveProxy::dump_config() { ESP_LOGCONFIG(TAG, "Z-Wave Proxy"); }
+void ZWaveProxy::dump_config() {
+  ESP_LOGCONFIG(TAG,
+                "Z-Wave Proxy:\n"
+                "  Home ID: %s",
+                format_hex_pretty(this->home_id_.data(), this->home_id_.size(), ':', false).c_str());
+}
 
 void ZWaveProxy::zwave_proxy_request(api::APIConnection *api_connection, api::enums::ZWaveProxyRequestType type) {
   switch (type) {
@@ -145,6 +158,17 @@ void ZWaveProxy::zwave_proxy_request(api::APIConnection *api_connection, api::en
   }
 }
 
+bool ZWaveProxy::set_home_id(const uint8_t *new_home_id) {
+  if (std::memcmp(this->home_id_.data(), new_home_id, this->home_id_.size()) == 0) {
+    ESP_LOGV(TAG, "Home ID unchanged");
+    return false;  // No change
+  }
+  std::memcpy(this->home_id_.data(), new_home_id, this->home_id_.size());
+  ESP_LOGI(TAG, "Home ID: %s", format_hex_pretty(this->home_id_.data(), this->home_id_.size(), ':', false).c_str());
+  this->home_id_ready_ = true;
+  return true;  // Home ID was changed
+}
+
 void ZWaveProxy::send_frame(const uint8_t *data, size_t length) {
   if (length == 1 && data[0] == this->last_response_) {
     ESP_LOGV(TAG, "Skipping sending duplicate response: 0x%02X", data[0]);
diff --git a/esphome/components/zwave_proxy/zwave_proxy.h b/esphome/components/zwave_proxy/zwave_proxy.h
index ea6837888b..a9123a81ca 100644
--- a/esphome/components/zwave_proxy/zwave_proxy.h
+++ b/esphome/components/zwave_proxy/zwave_proxy.h
@@ -56,6 +56,7 @@ class ZWaveProxy : public uart::UARTDevice, public Component {
   uint32_t get_home_id() {
     return encode_uint32(this->home_id_[0], this->home_id_[1], this->home_id_[2], this->home_id_[3]);
   }
+  bool set_home_id(const uint8_t *new_home_id);  // Store a new home ID. Returns true if it changed.
 
   void send_frame(const uint8_t *data, size_t length);
 
diff --git a/tests/components/spi/test.esp32-s3-ard.yaml b/tests/components/spi/test.esp32-s3-ard.yaml
new file mode 100644
index 0000000000..e4d4f20586
--- /dev/null
+++ b/tests/components/spi/test.esp32-s3-ard.yaml
@@ -0,0 +1,13 @@
+spi:
+  - id: three_spi
+    interface: spi3
+    clk_pin:
+      number: 47
+    mosi_pin:
+      number: 40
+  - id: hw_spi
+    interface: hardware
+    clk_pin:
+      number: 0
+    miso_pin:
+      number: 41