diff --git a/docs/requirements.md b/docs/requirements.md
index 8cfd5063..78718d4e 100644
--- a/docs/requirements.md
+++ b/docs/requirements.md
@@ -248,6 +248,7 @@ Internal architecture, components, and application logic.
 | DR-091 | Search result groups split TV into separate Shows and Episodes groups and add a People group (default order: Shows → Episodes → Movies → Songs → Albums → Artists → People); a stored `tvShows` order from before the split expands in place to shows+episodes so an upgrading user keeps their arrangement | UI | UR-060 | Done |
 | DR-092 | Video tap gestures resolve in `tapGestures.ts` (pure, unit-tested) rather than inline in `VideoPlayer.svelte`: `registerTap` returns `pending` for a first tap — the component defers `togglePlayPause` behind a `DOUBLE_TAP_WINDOW_MS` (300 ms) timer that a second tap cancels — or `seek` (+30 s right / −10 s left) for a second tap inside the window; a consumed second tap resets the state so a third tap starts fresh, and a swipe cancels the pending tap. The compatibility `click` the browser synthesizes after a touch tap is filtered in `handleVideoClick` so it cannot bypass the deferral. `resolveSeekTarget` converts the delta to the absolute position the facade requires, clamped per DR-095 and chained off a still-in-flight `pendingSeekTarget` so back-to-back skips accumulate instead of all resolving against a not-yet-updated position | UI | UR-061 | Done |
 | DR-094 | Frontend boundary tripwire (`scripts/check-frontend-boundary.sh`) detects Jellyfin item-type array literals **anywhere** in `src/` rather than only inline at an `includeItemTypes:` query site, so a category→type mapping cannot evade the check by being assigned to a named const (the evasion that let the `scoped-search` leak pass CI); requires two adjacent type literals so single-type presentation and `item.type ===` inspection stay legal, and caps the allowlist to force taxonomy into Rust instead of accumulating exceptions | Tooling | - | Done |
+| DR-097 | Transport authority (play/pause/toggle) lives in Rust for **webview-rendered** media, not just native. The controller tracks the state the HTML5 element reports (`html5_playing`, fed by `report_html5_state`, which now *stores* rather than only re-emitting); `play`/`pause`/`toggle_playback` consult it and drive the element by emitting a `ControlCommand` that `playerEvents.handleControlCommand` executes against the active adapter. A `stopped`/`idle` report clears it so the native backend (MPV/ExoPlayer) regains authority for music. The frontend facade no longer short-circuits transport into the adapter: `adapter.toggle()` previously decided play-vs-pause by reading `el.paused` off the DOM, a value that flips transiently while an element buffers or settles a seek — so two intents ~150 ms apart read *different* values, performed *opposing* actions, and self-sustained a play/pause loop needing no further input (observed on Android with a fully-buffered `readyState=4 networkState=1` element). Same "backend decides, adapter executes the primitive" split as `player_seek_video` | Player | UR-005 | Done |
 | DR-096 | `Html5PlayerAdapter.play()` is resilient to stall recovery: an in-flight attempt is memoised so concurrent callers (UI plus hls.js gap-controller recovery) share one `element.play()` instead of stacking calls, and an `AbortError` ("play() request was interrupted by a call to pause()") is logged at debug rather than pushed to `host.onError`. The browser raises it whenever a pending play promise is superseded by a pause/seek/source change, which hls.js does routinely while nudging past a stall — reporting it surfaced a player error roughly once per second for the whole stall and left the UI stuck showing paused | Player | UR-005 | Done |
 | DR-095 | Seek targets clamp strictly *inside* the media (`clampSeekTarget`, `END_SEEK_MARGIN_SECONDS` = 6 s ≈ one HLS segment) instead of to the exact `duration`. Landing on the duration makes hls.js request the segment whose start time lies past the end of the media (e.g. a 6330.324 s item → segment 1055 starting at 6336.33 s), which Jellyfin never produces; the fetch times out and hls.js' gap-controller stalls at the last buffered position, presenting as "unpausing or skipping bounces straight back to paused". Applied on both seek paths — the relative-skip `resolveSeekTarget` and the seek-bar drag, whose range input `max` is the duration itself — and floored at 0 so media shorter than the margin still seeks to the start | UI | UR-061 | Done |
 | DR-093 | Traceability coverage gate derives its requirement denominators from `requirements.md` at run time rather than hardcoded literals: `countDefinedRequirements` counts an ID only where it leads a markdown table row (ignoring the "Traces To" column and prose) and deduplicates IDs listed both in the definition tables and in the §3 traceability matrix; `computeCoverage` reports the *intersection* of traced and defined IDs so an ID traced in code but absent from `requirements.md` is surfaced as `orphaned` instead of inflating the ratio past 100%. UT/IT test identifiers are excluded as a separate taxonomy. CI and `bun run traces:coverage` share this computation and fail on both a sub-threshold and an impossible >100% result | Tooling | - | Done |
diff --git a/scripts/extract-traces.test.ts b/scripts/extract-traces.test.ts
index f65f7bf0..f7737102 100644
--- a/scripts/extract-traces.test.ts
+++ b/scripts/extract-traces.test.ts
@@ -175,8 +175,8 @@ describe("live requirements.md", () => {
 
     expect(defined.UR).toBe(61);
     expect(defined.IR).toBe(29);
-    expect(defined.DR).toBe(93);
+    expect(defined.DR).toBe(94);
     expect(defined.JA).toBe(32);
-    expect(defined.total).toBe(215);
+    expect(defined.total).toBe(216);
   });
 });
diff --git a/src-tauri/src/player/mod.rs b/src-tauri/src/player/mod.rs
index a5857ab3..c2b9775d 100644
--- a/src-tauri/src/player/mod.rs
+++ b/src-tauri/src/player/mod.rs
@@ -152,6 +152,17 @@ pub struct PlayerController {
 
     // Auto-play episode counter (session-based, resets on manual play)
     autoplay_episode_count: Arc<Mutex<u32>>,
+
+    // Last state reported by a webview-rendered HTML5 <video>/<audio> element.
+    //
+    // Webview-rendered media is played by an element the native backend cannot
+    // reach, so the backend's own state() says nothing about it. Tracking the
+    // REPORTED state here is what lets transport (play/pause/toggle) be decided
+    // in Rust for that media instead of the frontend reading `el.paused` off the
+    // DOM — a value that flips transiently while buffering/seeking and caused
+    // competing intents to take opposing actions. `None` means no webview media
+    // is active and the native backend is authoritative. See DR-097.
+    html5_playing: Arc<Mutex<Option<bool>>>,
 }
 
 impl PlayerController {
@@ -174,6 +185,7 @@ impl PlayerController {
             position_throttler,
             end_reason: Arc::new(Mutex::new(None)),
             autoplay_episode_count: Arc::new(Mutex::new(0)),
+            html5_playing: Arc::new(Mutex::new(None)),
         };
 
         // Start background timer thread for sleep timer countdown
@@ -476,21 +488,72 @@ impl PlayerController {
         Ok(())
     }
 
+    /// True while webview-rendered media (HTML5 `<video>`/`<audio>`) is the real
+    /// player, so transport must be routed to it rather than the native backend.
+    ///
+    /// TRACES: UR-005 | DR-097
+    pub fn is_html5_active(&self) -> bool {
+        self.html5_playing.lock_safe().is_some()
+    }
+
+    /// Whether the webview element last reported itself as playing. Meaningless
+    /// unless [`Self::is_html5_active`] is true.
+    ///
+    /// TRACES: UR-005 | DR-097
+    pub fn html5_is_playing(&self) -> bool {
+        self.html5_playing.lock_safe().unwrap_or(false)
+    }
+
+    /// Send a transport intent to the webview element that is rendering media.
+    fn emit_html5_control(&self, action: &str) {
+        if let Some(emitter) = self.event_emitter.lock_safe().as_ref() {
+            emitter.emit(PlayerStatusEvent::ControlCommand {
+                action: action.to_string(),
+                position: None,
+            });
+        }
+    }
+
     /// Play/resume playback
     pub fn play(&self) -> Result<(), PlayerError> {
         debug!("[PlayerController] play");
+        // Webview-rendered media: the native backend isn't playing it, so drive
+        // the element via a ControlCommand instead (DR-097).
+        if self.is_html5_active() {
+            self.emit_html5_control("play");
+            return Ok(());
+        }
         let mut backend = self.backend.lock_safe();
         backend.play()
     }
 
     /// Pause playback
     pub fn pause(&self) -> Result<(), PlayerError> {
+        if self.is_html5_active() {
+            self.emit_html5_control("pause");
+            return Ok(());
+        }
         let mut backend = self.backend.lock_safe();
         backend.pause()
     }
 
-    /// Toggle play/pause
+    /// Toggle play/pause.
+    ///
+    /// The decision is made HERE, from authoritative state — the reported webview
+    /// state for HTML5-rendered media, or the native backend's state otherwise.
+    /// The frontend must never decide this from the DOM (see DR-097).
+    ///
+    /// TRACES: UR-005 | DR-097
     pub fn toggle_playback(&self) -> Result<(), PlayerError> {
+        if self.is_html5_active() {
+            let action = if self.html5_is_playing() {
+                "pause"
+            } else {
+                "play"
+            };
+            self.emit_html5_control(action);
+            return Ok(());
+        }
         let mut backend = self.backend.lock_safe();
         if backend.state().is_playing() {
             backend.pause()
@@ -890,6 +953,23 @@ impl PlayerController {
     /// Re-emits a `StateChanged` event identical to what MpvBackend/ExoPlayer
     /// would emit, so `playerEvents.ts` needs no HTML5-specific branch.
     pub fn report_html5_state(&self, state: String, media_id: Option<String>) {
+        // Track it: this is the authoritative play/pause state for
+        // webview-rendered media, and what transport decisions read (DR-097).
+        // "stopped"/"idle" mean the element is gone, so hand authority back to
+        // the native backend — otherwise music playback would keep emitting
+        // ControlCommands at a element that no longer exists.
+        {
+            let mut tracked = self.html5_playing.lock_safe();
+            *tracked = match state.as_str() {
+                "playing" => Some(true),
+                // "loading" counts as active-but-not-playing so a toggle during
+                // load resolves to "play" rather than falling through to the
+                // native backend.
+                "paused" | "loading" => Some(false),
+                // "stopped"/"idle": element is gone, native backend resumes authority.
+                _ => None,
+            };
+        }
         if let Some(emitter) = self.event_emitter.lock_safe().as_ref() {
             emitter.emit(PlayerStatusEvent::StateChanged { state, media_id });
         }
@@ -1489,6 +1569,156 @@ mod tests {
         }
     }
 
+    // ===== HTML5 transport authority (DR-097) =====
+    //
+    // Webview-rendered video is played by an element the native backend cannot
+    // reach, so transport for it must be decided from the state the element
+    // REPORTS and executed by emitting a ControlCommand. Previously the frontend
+    // decided play-vs-pause itself by reading `el.paused` off the DOM, which
+    // flips transiently while buffering/seeking — two intents ~150ms apart read
+    // different values, took opposing actions, and self-sustained a pause loop.
+
+    #[test]
+    fn test_html5_state_is_tracked_from_reports() {
+        let controller = PlayerController::default();
+        let emitter = Arc::new(CapturingEmitter::new());
+        controller.set_event_emitter(emitter.clone());
+
+        // No HTML5 media reported yet: the native backend stays authoritative.
+        assert!(!controller.is_html5_active());
+
+        controller.report_html5_state("playing".to_string(), Some("item-1".to_string()));
+        assert!(controller.is_html5_active());
+        assert!(controller.html5_is_playing());
+
+        controller.report_html5_state("paused".to_string(), Some("item-1".to_string()));
+        assert!(controller.is_html5_active());
+        assert!(!controller.html5_is_playing());
+    }
+
+    #[test]
+    fn test_html5_toggle_from_paused_emits_play_control() {
+        let controller = PlayerController::default();
+        let emitter = Arc::new(CapturingEmitter::new());
+        controller.set_event_emitter(emitter.clone());
+        controller.report_html5_state("paused".to_string(), Some("item-1".to_string()));
+
+        controller.toggle_playback().unwrap();
+
+        let controls: Vec<_> = emitter
+            .events()
+            .into_iter()
+            .filter_map(|e| match e {
+                PlayerStatusEvent::ControlCommand { action, .. } => Some(action),
+                _ => None,
+            })
+            .collect();
+        assert_eq!(controls, vec!["play".to_string()]);
+    }
+
+    #[test]
+    fn test_html5_toggle_from_playing_emits_pause_control() {
+        let controller = PlayerController::default();
+        let emitter = Arc::new(CapturingEmitter::new());
+        controller.set_event_emitter(emitter.clone());
+        controller.report_html5_state("playing".to_string(), Some("item-1".to_string()));
+
+        controller.toggle_playback().unwrap();
+
+        let controls: Vec<_> = emitter
+            .events()
+            .into_iter()
+            .filter_map(|e| match e {
+                PlayerStatusEvent::ControlCommand { action, .. } => Some(action),
+                _ => None,
+            })
+            .collect();
+        assert_eq!(controls, vec!["pause".to_string()]);
+    }
+
+    #[test]
+    fn test_html5_repeated_toggles_alternate_and_never_repeat_an_action() {
+        // The loop signature: two intents in quick succession must NOT both
+        // resolve the same way, and must not produce opposing actions from a
+        // stale read. Rust's own tracked state makes the sequence deterministic
+        // as long as the element reports back between intents.
+        let controller = PlayerController::default();
+        let emitter = Arc::new(CapturingEmitter::new());
+        controller.set_event_emitter(emitter.clone());
+        controller.report_html5_state("playing".to_string(), Some("item-1".to_string()));
+
+        controller.toggle_playback().unwrap();
+        // Element confirms the pause it was told to do.
+        controller.report_html5_state("paused".to_string(), Some("item-1".to_string()));
+        controller.toggle_playback().unwrap();
+
+        let controls: Vec<_> = emitter
+            .events()
+            .into_iter()
+            .filter_map(|e| match e {
+                PlayerStatusEvent::ControlCommand { action, .. } => Some(action),
+                _ => None,
+            })
+            .collect();
+        assert_eq!(controls, vec!["pause".to_string(), "play".to_string()]);
+    }
+
+    #[test]
+    fn test_html5_play_and_pause_emit_control_commands() {
+        let controller = PlayerController::default();
+        let emitter = Arc::new(CapturingEmitter::new());
+        controller.set_event_emitter(emitter.clone());
+        controller.report_html5_state("paused".to_string(), Some("item-1".to_string()));
+
+        controller.play().unwrap();
+        controller.pause().unwrap();
+
+        let controls: Vec<_> = emitter
+            .events()
+            .into_iter()
+            .filter_map(|e| match e {
+                PlayerStatusEvent::ControlCommand { action, .. } => Some(action),
+                _ => None,
+            })
+            .collect();
+        assert_eq!(controls, vec!["play".to_string(), "pause".to_string()]);
+    }
+
+    #[test]
+    fn test_html5_stopped_report_releases_transport_to_native_backend() {
+        // When webview video goes away, transport must fall back to the native
+        // backend (music playback must not keep emitting ControlCommands).
+        let controller = PlayerController::default();
+        let emitter = Arc::new(CapturingEmitter::new());
+        controller.set_event_emitter(emitter.clone());
+
+        controller.report_html5_state("playing".to_string(), Some("item-1".to_string()));
+        assert!(controller.is_html5_active());
+
+        controller.report_html5_state("stopped".to_string(), None);
+        assert!(!controller.is_html5_active());
+    }
+
+    #[test]
+    fn test_html5_transport_emits_exactly_one_control_per_intent() {
+        // Guards against a double-drive on platforms where the *backend* is also
+        // webview-based (WebviewAudioBackend on Windows): the html5 short-circuit
+        // must replace the backend call, not run in addition to it.
+        let controller = PlayerController::default();
+        let emitter = Arc::new(CapturingEmitter::new());
+        controller.set_event_emitter(emitter.clone());
+        controller.report_html5_state("playing".to_string(), Some("item-1".to_string()));
+
+        controller.pause().unwrap();
+
+        let controls = emitter
+            .events()
+            .into_iter()
+            .filter(|e| matches!(e, PlayerStatusEvent::ControlCommand { .. }))
+            .count();
+        assert_eq!(controls, 1, "one intent must produce exactly one control");
+    }
+
     #[test]
     fn test_controller_volume_default() {
         let controller = PlayerController::default();
diff --git a/src/lib/player/index.ts b/src/lib/player/index.ts
index e4dee1c2..49a50d2b 100644
--- a/src/lib/player/index.ts
+++ b/src/lib/player/index.ts
@@ -12,7 +12,7 @@
  * derived + merged (remote-session-aware) stores so UI can import state and
  * actions from one place, in both local and remote modes.
  *
- * TRACES: UR-005 | DR-001, DR-009
+ * TRACES: UR-005 | DR-001, DR-009, DR-097 | UT-089
  */
 
 import { get } from "svelte/store";
@@ -83,18 +83,29 @@ function requireHandle(): string {
 // Transport controls (no repository handle required)
 // ---------------------------------------------------------------------------
 
+// Transport intents ALWAYS go to the backend, in both native and HTML5 modes.
+//
+// These used to short-circuit into the active video adapter, which made the
+// webview the decider: `adapter.toggle()` read `el.paused` off the DOM and
+// flipped the element, so Rust never saw the intent. `el.paused` flips
+// transiently while an element buffers or settles a seek, so two intents
+// ~150ms apart could read different values and take opposing actions — a
+// self-sustaining play/pause loop.
+//
+// Now Rust decides from PlayerController state and drives the element back
+// through a `ControlCommand` event (handled in playerEvents.ts), the same
+// "backend decides, adapter executes the primitive" split used by
+// player_seek_video. Do NOT reintroduce an adapter short-circuit here.
+
 async function play() {
-  if (activeAdapter) return void (await activeAdapter.play());
   await commands.playerPlay();
 }
 
 async function pause() {
-  if (activeAdapter) return void (await activeAdapter.pause());
   await commands.playerPause();
 }
 
 async function toggle() {
-  if (activeAdapter) return void (await activeAdapter.toggle());
   await commands.playerToggle();
 }
 
diff --git a/src/lib/player/transport.authority.test.ts b/src/lib/player/transport.authority.test.ts
new file mode 100644
index 00000000..716cf389
--- /dev/null
+++ b/src/lib/player/transport.authority.test.ts
@@ -0,0 +1,113 @@
+/**
+ * Transport authority: play/pause/toggle are DECIDED in Rust, never in the webview.
+ *
+ * TRACES: UR-005 | DR-097 | UT-089
+ *
+ * The frontend used to short-circuit transport controls whenever a video adapter
+ * was registered: `toggle()` read `el.paused` off the DOM and flipped the element
+ * directly, so the Rust `PlayerController` never saw the intent and could not
+ * serialise competing ones. Because `el.paused` flips transiently while an HTML5
+ * element buffers or settles a seek, two intents arriving ~150ms apart could read
+ * *different* values and perform *opposing* actions — one playing, one pausing —
+ * which is the self-sustaining play/pause loop observed on Android.
+ *
+ * The rule these tests pin: a transport intent always reaches the backend. Rust
+ * decides play-vs-pause from controller state and drives the webview element back
+ * through a ControlCommand event (the same "backend decides, adapter executes"
+ * split `player_seek_video` already uses).
+ */
+
+import { describe, it, expect, vi, beforeEach } from "vitest";
+
+const mockCommands = {
+  playerPlay: vi.fn(async () => ({})),
+  playerPause: vi.fn(async () => ({})),
+  playerToggle: vi.fn(async () => ({})),
+  playerStop: vi.fn(async () => ({})),
+};
+
+vi.mock("$lib/api/bindings", () => ({
+  commands: mockCommands,
+  // Stores pulled in transitively subscribe to typed events at module load.
+  events: {
+    playerStatusEvent: { listen: vi.fn(async () => () => {}) },
+    downloadEvent: { listen: vi.fn(async () => () => {}) },
+    searchEvent: { listen: vi.fn(async () => () => {}) },
+  },
+}));
+
+vi.mock("$lib/stores/auth", () => ({
+  auth: {
+    subscribe: (fn: (v: unknown) => void) => {
+      fn({ isAuthenticated: true });
+      return () => {};
+    },
+    getRepository: () => ({ getHandle: () => "handle-1" }),
+  },
+}));
+
+/** A video adapter that records whether the facade reached into it directly. */
+function makeAdapter() {
+  return {
+    kind: "html5" as const,
+    play: vi.fn(async () => {}),
+    pause: vi.fn(async () => {}),
+    toggle: vi.fn(async () => true),
+    seekElement: vi.fn(async () => {}),
+    reloadSource: vi.fn(async () => {}),
+    attach: vi.fn(),
+    dispose: vi.fn(async () => {}),
+    setVolume: vi.fn(),
+    setMuted: vi.fn(),
+    selectSubtitle: vi.fn(async () => {}),
+    getPosition: vi.fn(() => 0),
+    load: vi.fn(async () => {}),
+  };
+}
+
+describe("transport authority lives in Rust", () => {
+  let playerController: any;
+  let adapter: ReturnType<typeof makeAdapter>;
+
+  beforeEach(async () => {
+    vi.clearAllMocks();
+    vi.resetModules();
+    ({ playerController } = await import("./index"));
+    adapter = makeAdapter();
+    playerController.setActiveAdapter(adapter);
+  });
+
+  it("routes toggle to the backend even when a video adapter is active", async () => {
+    await playerController.toggle();
+
+    expect(mockCommands.playerToggle).toHaveBeenCalledTimes(1);
+    // The webview must NOT decide play-vs-pause from the DOM.
+    expect(adapter.toggle).not.toHaveBeenCalled();
+  });
+
+  it("routes play to the backend even when a video adapter is active", async () => {
+    await playerController.play();
+
+    expect(mockCommands.playerPlay).toHaveBeenCalledTimes(1);
+    expect(adapter.play).not.toHaveBeenCalled();
+  });
+
+  it("routes pause to the backend even when a video adapter is active", async () => {
+    await playerController.pause();
+
+    expect(mockCommands.playerPause).toHaveBeenCalledTimes(1);
+    expect(adapter.pause).not.toHaveBeenCalled();
+  });
+
+  it("still routes transport to the backend with no adapter (audio path unchanged)", async () => {
+    playerController.clearActiveAdapter();
+
+    await playerController.toggle();
+    await playerController.play();
+    await playerController.pause();
+
+    expect(mockCommands.playerToggle).toHaveBeenCalledTimes(1);
+    expect(mockCommands.playerPlay).toHaveBeenCalledTimes(1);
+    expect(mockCommands.playerPause).toHaveBeenCalledTimes(1);
+  });
+});
diff --git a/src/lib/services/playerEvents.ts b/src/lib/services/playerEvents.ts
index 8f68aef3..21b94b58 100644
--- a/src/lib/services/playerEvents.ts
+++ b/src/lib/services/playerEvents.ts
@@ -5,7 +5,7 @@
  * frontend stores accordingly. This enables push-based updates instead
  * of polling.
  *
- * TRACES: UR-005, UR-019, UR-023, UR-026 | DR-001, DR-028, DR-047
+ * TRACES: UR-005, UR-019, UR-023, UR-026 | DR-001, DR-028, DR-047, DR-097
  */
 
 import { type UnlistenFn } from "@tauri-apps/api/event";
@@ -306,9 +306,15 @@ function handleSleepTimerChanged(mode: SleepTimerMode, remainingSeconds: number)
 
 /**
  * Route a backend-originated control command to the active player adapter, so a
- * backend intent (lockscreen/remote/sleep) can drive the webview <video> element
- * that Rust cannot reach directly. No-op when no video adapter is active (audio
- * playback is already fully backend-driven).
+ * backend intent can drive the webview <video>/<audio> element that Rust cannot
+ * reach directly. No-op when no adapter is active (native playback is already
+ * fully backend-driven).
+ *
+ * This is the EXECUTION half of transport authority: for webview-rendered media
+ * the Rust controller decides play-vs-pause from the state the element reported
+ * and emits it here as a ControlCommand. UI intents go *to* the backend (see the
+ * facade in $lib/player) and come back through this path — never short-circuited
+ * in the webview, which is what caused the DR-097 pause loop.
  */
 function handleControlCommand(action: string, position: number | null): void {
   const adapter = playerController.getActiveAdapter();