diff --git a/tdisplay_p4/main/CMakeLists.txt b/tdisplay_p4/main/CMakeLists.txt
index 2d4ccde..2b47f75 100644
--- a/tdisplay_p4/main/CMakeLists.txt
+++ b/tdisplay_p4/main/CMakeLists.txt
@@ -19,6 +19,14 @@ idf_component_register(
 		c6_at              # AT-over-SDIO Wi-Fi/BLE driver for the on-board ESP32-C6 (P4-only)
 	INCLUDE_DIRS
 		"."
+		# The repo's own src/ MUST come before the meshcore component's include dir. The vendored
+		# core publishes core/src publicly and still carries OLD copies of the headers whose .cpp
+		# files it deliberately excludes (SdNvsPrefs, TouchPrefsStore, ...) because our src/ ships
+		# them. Without this line an angle include here resolves to the stale core header while the
+		# linker binds the new object -- so anything added to those headers since the core tag looks
+		# like it does not exist (SdNvsPrefs::useFile, touchPrefsReload). Same headers the app
+		# component compiles against, so main.cpp and the app layer now agree.
+		"${CMAKE_CURRENT_LIST_DIR}/../../src"
 )
 
 # The vendored MeshCore core + Arduino libs trip IDF's blanket -Werror=all. A bare
diff --git a/tdisplay_p4/main/main.cpp b/tdisplay_p4/main/main.cpp
index 8a7f393..4bb78d1 100644
--- a/tdisplay_p4/main/main.cpp
+++ b/tdisplay_p4/main/main.cpp
@@ -21,6 +21,7 @@
 #include <helpers/esp32/MultiTransportCompanionInterface.h>
 #include <helpers/esp32/WifiRuntimeStore.h>   // wifiConfig* (runtime Wi-Fi state the loop drives)
 #include <helpers/esp32/TouchPrefsStore.h>    // touchPrefsBuildLocalTz + WIFI_CONFIG_* sizes
+#include <helpers/esp32/SdNvsPrefs.h>        // route prefs off internal-flash NVS onto the SD card
 #include "esp_hosted.h"
 #include "esp_netif.h"
 #include "esp_event.h"
@@ -212,6 +213,30 @@ static void wadameshSetup() {
     }
     store.useSdMmcStorage();
   }
+
+  // Route the touch settings + Wi-Fi creds off internal-flash NVS onto the SD card, the same way
+  // every S3 board does (src/main.cpp). This board never did it, so although contacts and chat
+  // history correctly moved to SD above, EVERY pref write still hit NVS in internal flash.
+  //
+  // That matters here beyond wear: this panel is a DSI video one whose framebuffer is streamed
+  // continuously out of PSRAM, and an internal-flash write visibly disturbs it -- a whole-screen
+  // flash, reported as #167. The trigger list is exactly the pref/thread-state writes: a message
+  // arriving, opening a chat that has unread messages, and deleting a contact (which writes
+  // repeatedly and so flickers for the whole operation). Writes that go to the SD card instead
+  // never do it -- panning the map downloads and writes tiles continuously with no flicker at all,
+  // which is what pinned this down.
+  //
+  // Only when a card is actually mounted: with no card there is nowhere better to put prefs, and
+  // moving them from NVS to the internal FAT partition would trade one flash write for another
+  // while needlessly changing where existing users' settings live. Old NVS values still load and
+  // migrate on their next save, so this is transparent. touchPrefsReload() is required because
+  // boot already read prefs (UI rotation) through the old backend and cached the blob.
+  if (g_sd_ok) {
+    SdNvsPrefs::useFile((fs::FS*)&SD_MMC, "/meshcomod");
+    touchPrefsReload();
+    printf("[storage] prefs -> SD /meshcomod (off internal-flash NVS)\n");
+  }
+
   store.begin();
 
   the_mesh.begin(disp != NULL);