diff --git a/static/boot.js b/static/boot.js
index 2d50695a..1fffefa8 100644
--- a/static/boot.js
+++ b/static/boot.js
@@ -855,6 +855,7 @@ window._micPendingSend=window._micPendingSend||false;
   let _voiceModeThinkingSid=null;
   let _browserTtsKeepAlive=null;
   let _browserTtsWatchdog=null;
+  let _browserTtsSuppressNextErrorRearm=false;
   const SILENCE_MS=1800; // auto-send after 1.8s silence
 
   function _clearBrowserTtsRecovery(){
@@ -870,11 +871,13 @@ window._micPendingSend=window._micPendingSend||false;
 
   function _armBrowserTtsRecovery(clean, rate){
     _clearBrowserTtsRecovery();
+    _browserTtsSuppressNextErrorRearm=false;
     const safeRate=(Number.isFinite(rate)&&rate>0)?rate:1;
     // Chromium can drop utter.onend on later turns, so force a recovery path.
     const watchdogMs=Math.max(4000,Math.round((String(clean||'').length/(12*safeRate))*1000)+10000);
     _browserTtsWatchdog=setTimeout(()=>{
       if(!_voiceModeActive||_voiceModeState!=='speaking') return;
+      _browserTtsSuppressNextErrorRearm=true;
       try{ speechSynthesis.cancel(); }catch(_){}
       _clearBrowserTtsRecovery();
       _startListening();
@@ -1095,12 +1098,17 @@ window._micPendingSend=window._micPendingSend||false;
     if(!isNaN(savedPitch)) utter.pitch=Math.min(2,Math.max(0,savedPitch));
 
     utter.onend=()=>{
+      _browserTtsSuppressNextErrorRearm=false;
       _clearBrowserTtsRecovery();
       // After speaking, go back to listening
       if(_voiceModeActive) setTimeout(()=>_startListening(),500);
     };
     utter.onerror=()=>{
       _clearBrowserTtsRecovery();
+      if(_browserTtsSuppressNextErrorRearm){
+        _browserTtsSuppressNextErrorRearm=false;
+        return;
+      }
       if(_voiceModeActive) setTimeout(()=>_startListening(),1000);
     };
 
@@ -1167,6 +1175,7 @@ window._micPendingSend=window._micPendingSend||false;
     _voiceModeActive=false;
     _voiceModeState='idle';
     _voiceModeThinkingSid=null;
+    _browserTtsSuppressNextErrorRearm=false;
     modeBtn.classList.remove('active');
     _setButtonTooltip(modeBtn, t('voice_mode_toggle'));
     bar.style.display='none';
diff --git a/tests/test_issue3983_browser_tts_watchdog.py b/tests/test_issue3983_browser_tts_watchdog.py
index 69e73387..03e55bfd 100644
--- a/tests/test_issue3983_browser_tts_watchdog.py
+++ b/tests/test_issue3983_browser_tts_watchdog.py
@@ -9,6 +9,7 @@ def test_boot_js_declares_browser_tts_recovery_helpers():
     src = (REPO / "static" / "boot.js").read_text(encoding="utf-8")
     assert "let _browserTtsKeepAlive=null;" in src
     assert "let _browserTtsWatchdog=null;" in src
+    assert "let _browserTtsSuppressNextErrorRearm=false;" in src
     assert "function _clearBrowserTtsRecovery()" in src
     assert "function _armBrowserTtsRecovery(clean, rate)" in src
 
@@ -20,6 +21,7 @@ def test_browser_tts_watchdog_rearms_listening_if_onend_drops():
     arm_body = src[arm_idx : arm_idx + 1400]
     assert "_browserTtsWatchdog=setTimeout" in arm_body
     assert "_voiceModeState!=='speaking'" in arm_body
+    assert "_browserTtsSuppressNextErrorRearm=true;" in arm_body
     assert "speechSynthesis.cancel()" in arm_body
     assert "_startListening();" in arm_body
     assert "_browserTtsKeepAlive=setInterval" in arm_body
@@ -37,6 +39,8 @@ def test_browser_tts_callbacks_and_deactivate_clear_recovery_handles():
     assert speak_body.count("_clearBrowserTtsRecovery();") >= 2, (
         "Both browser TTS completion callbacks must clear watchdog/keep-alive handles."
     )
+    assert "_browserTtsSuppressNextErrorRearm=false;" in speak_body
+    assert "if(_browserTtsSuppressNextErrorRearm){" in speak_body
     assert "_armBrowserTtsRecovery(clean, utter.rate);" in speak_body
 
     deactivate_match = re.search(
@@ -49,6 +53,7 @@ def test_browser_tts_callbacks_and_deactivate_clear_recovery_handles():
     assert "_clearBrowserTtsRecovery();" in deactivate_body, (
         "_deactivate() must clear browser TTS watchdog/keep-alive handles."
     )
+    assert "_browserTtsSuppressNextErrorRearm=false;" in deactivate_body
 
 
 def test_edge_audio_branch_stays_separate():