feat(core,host,client): PyroWave datagram-aligned packets + partial-frame delivery (Phase 4, §4.4)

PyroWave AUs now packetize on the negotiated shard payload, so a lost datagram
costs a few wavelet blocks of localized blur rather than a whole frame — and the
client can render an aged-out lossy frame instead of freezing until the next one.

Host (opt-in, PyroWave only):
- The encoder packetizes at the shard payload behind a 4-byte window prefix
  (used-len u16 + kind u16). Whole packets pack into WIN_PACKED windows; a packet
  too large for one shard (PyroWave 32x32 blocks are atomic and can exceed a
  shard) rides a WIN_FRAG_FIRST/CONT/LAST chain. `set_wire_chunking()` joins the
  Encoder trait (forwarded through TrackedEncoder — the silent-no-op trap);
  EncodedFrame.chunk_aligned marks the AU.
- virtual_stream tags the AU with USER_FLAG_CHUNK_ALIGNED and re-applies chunking
  after every encoder (re)build, the adaptive-bitrate rebuild included.

Core:
- USER_FLAG_CHUNK_ALIGNED (0x40) wire bit. Reassembler opt-in
  (set_deliver_partial): a chunk-aligned frame that ages out with holes is handed
  over as Frame{complete:false} — received shards at their exact offsets, missing
  ranges zero-filled — instead of being dropped. Partials age out on a tight 30ms
  fuse (PARTIAL_WINDOW_NS) instead of the 120ms loss window: each frame is
  independently decodable, so an ancient partial has no value in a live stream.
  Newest-wins. A partial still counts as dropped for loss reporting.

Client (PyroWave decode):
- The session opts in when codec == PyroWave. The decoder walks the AU
  window-by-window, skipping zero (missing) windows and reassembling FRAG chains,
  then decodes whatever survived. A newest-decoded-index guard drops partials the
  pump has already moved past (no time-travel present).

Also fixes a redundant-closure clippy nit in the PyroWave planar-present path.

Validated on an RTX 5070 Ti under 2% netem loss with FEC pinned off: 60fps
sustained entirely via partials, e2e 43ms p50 (146ms before the fuse) vs 23ms
lossless, no keyframe-recovery chatter. Tests green: core 149, host 310 + the
GPU-gated encoder smoke (framed-window walk + FRAG reassembly + upstream
round-trip), client 26; clippy clean on the pyrowave feature combos.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

crates/pf-client-core/src/session.rs

@@ -289,7 +289,12 @@ fn pump(
     let built = if connector.codec == punktfunk_core::quic::CODEC_PYROWAVE {
         let mode = connector.mode();
         match params.vulkan.as_ref() {
-            Some(vk) => Decoder::new_pyrowave(vk, mode.width, mode.height),
+            Some(vk) => Decoder::new_pyrowave(
+                vk,
+                mode.width,
+                mode.height,
+                connector.shard_payload as usize,
+            ),
             None => Err(anyhow::anyhow!(
                 "pyrowave session without a presenter device"
             )),
@@ -324,6 +329,8 @@ fn pump(
     // step, drift) keep the capture-clock latency stats honest — never cached at session start.
     let clock_offset_live = connector.clock_offset_shared();
     let mut total_frames = 0u64;
+    // Newest frame index handed to the decoder — the staleness bar for late partials.
+    let mut newest_decoded_idx: Option<u32> = None;
     let mut window_start = Instant::now();
     let mut frames_n = 0u32;
     let mut bytes_n = 0u64;
@@ -432,7 +439,21 @@ fn pump(
                     }
                     None => next_expected_index = Some(frame.frame_index.wrapping_add(1)),
                 }
-                match decoder.decode(&frame.data) {
+                // A PARTIAL that lost the race (a newer frame already decoded) is pure
+                // time travel — skip it; each PyroWave frame is independent, so nothing
+                // downstream needs it. Completes keep the normal path (reorder is handled
+                // by the continuity gate).
+                if !frame.complete
+                    && newest_decoded_idx
+                        .is_some_and(|n: u32| n.wrapping_sub(frame.frame_index) <= u32::MAX / 2)
+                {
+                    continue;
+                }
+                newest_decoded_idx = Some(match newest_decoded_idx {
+                    Some(n) if frame.frame_index.wrapping_sub(n) > u32::MAX / 2 => n,
+                    _ => frame.frame_index,
+                });
+                match decoder.decode_frame(&frame.data, frame.flags, frame.complete) {
                     Ok(Some(image)) => {
                         // Fold this decoded frame through the shared freeze gate: it reads the AU's
                         // re-anchor wire flags (FLAG_SOF IDR marker / RECOVERY_ANCHOR / RECOVERY_POINT),

crates/pf-client-core/src/video.rs

@@ -572,10 +572,18 @@ impl Decoder {
     /// compute on the presenter's device, no FFmpeg. `codec_id` is irrelevant (kept as
     /// HEVC so an — impossible — demotion path stays well-formed).
     #[cfg(all(target_os = "linux", feature = "pyrowave"))]
-    pub fn new_pyrowave(vk: &VulkanDecodeDevice, width: u32, height: u32) -> Result<Decoder> {
+    pub fn new_pyrowave(
+        vk: &VulkanDecodeDevice,
+        width: u32,
+        height: u32,
+        shard_payload: usize,
+    ) -> Result<Decoder> {
         Ok(Decoder {
             backend: Backend::PyroWave(Box::new(crate::video_pyrowave::PyroWaveDecoder::new(
-                vk, width, height,
+                vk,
+                width,
+                height,
+                shard_payload,
             )?)),
             codec_id: ffmpeg::codec::Id::HEVC,
             vaapi_fails: 0,
@@ -610,8 +618,24 @@ impl Decoder {
     /// pump asks the host for a fresh IDR — under the infinite GOP nothing else resyncs a
     /// rebuilt/erroring decoder, so skipping this leaves the picture gray/frozen for good.
     pub fn decode(&mut self, au: &[u8]) -> Result<Option<DecodedImage>> {
+        self.decode_frame(au, 0, true)
+    }
+
+    /// [`decode`](Self::decode) with the AU's wire facts: `user_flags` (chunk-aligned AUs
+    /// are parsed in shard windows — [`punktfunk_core::packet::USER_FLAG_CHUNK_ALIGNED`])
+    /// and completeness (`false` = a partial delivery; only the PyroWave backend decodes
+    /// those — as one frame of localized blur, plan §4.4).
+    pub fn decode_frame(
+        &mut self,
+        au: &[u8],
+        user_flags: u32,
+        complete: bool,
+    ) -> Result<Option<DecodedImage>> {
         let result = match &mut self.backend {
-            Backend::Vulkan(v) => v.decode(au).map(|f| f.map(DecodedImage::VkFrame)),
+            Backend::Vulkan(v) => {
+                debug_assert!(complete, "partial AUs are pyrowave-only");
+                v.decode(au).map(|f| f.map(DecodedImage::VkFrame))
+            }
             #[cfg(target_os = "linux")]
             Backend::Vaapi(v) => v.decode(au).map(|f| f.map(DecodedImage::Dmabuf)),
             #[cfg(windows)]
@@ -620,7 +644,12 @@ impl Decoder {
             // error; the pump surfaces it and the session falls back to HEVC by
             // renegotiation (plan §4.6), not by decoder swap.
             #[cfg(all(target_os = "linux", feature = "pyrowave"))]
-            Backend::PyroWave(p) => return Ok(p.decode(au)?.map(DecodedImage::PyroWave)),
+            Backend::PyroWave(p) => {
+                let aligned = user_flags & punktfunk_core::packet::USER_FLAG_CHUNK_ALIGNED != 0;
+                return Ok(p
+                    .decode_frame(au, aligned, complete)?
+                    .map(DecodedImage::PyroWave));
+            }
             Backend::Software(s) => return Ok(s.decode(au)?.map(DecodedImage::Cpu)),
         };
         match result {

crates/pf-client-core/src/video_pyrowave.rs

@@ -190,6 +190,9 @@ pub struct PyroWaveDecoder {
     fence: vk::Fence,
     width: u32,
     height: u32,
+    /// The wire shard payload — the parse-window size for chunk-aligned AUs (§4.4): each
+    /// window holds whole self-delimiting codec packets, zero-padded to the window.
+    wire_window: usize,
 }
 
 // SAFETY: used only from the single decode thread; the shared-queue accesses go through
@@ -197,7 +200,12 @@ pub struct PyroWaveDecoder {
 unsafe impl Send for PyroWaveDecoder {}
 
 impl PyroWaveDecoder {
-    pub fn new(vkd: &VulkanDecodeDevice, width: u32, height: u32) -> Result<PyroWaveDecoder> {
+    pub fn new(
+        vkd: &VulkanDecodeDevice,
+        width: u32,
+        height: u32,
+        shard_payload: usize,
+    ) -> Result<PyroWaveDecoder> {
         if !vkd.pyrowave_decode {
             bail!("presenter device lacks the PyroWave compute feature set");
         }
@@ -207,13 +215,14 @@ impl PyroWaveDecoder {
         // SAFETY: the handles in `vkd` are the presenter's live instance/device (it
         // outlives the decoder — same contract the FFmpeg Vulkan backend relies on);
         // `Hold` pins the reconstructed create-infos for the pyrowave device's lifetime.
-        unsafe { Self::new_inner(vkd, width, height) }
+        unsafe { Self::new_inner(vkd, width, height, shard_payload) }
     }
 
     unsafe fn new_inner(
         vkd: &VulkanDecodeDevice,
         width: u32,
         height: u32,
+        shard_payload: usize,
     ) -> Result<PyroWaveDecoder> {
         let static_fn = ash::StaticFn {
             get_instance_proc_addr: std::mem::transmute::<usize, vk::PFN_vkGetInstanceProcAddr>(
@@ -380,25 +389,113 @@ impl PyroWaveDecoder {
             fence,
             width,
             height,
+            wire_window: shard_payload.max(64),
         })
     }
 
-    /// One AU in → one frame out (the AU is a complete pyrowave frame: one packet).
+    /// One AU in → one frame out. `aligned` = the AU is shard-window chunked (each
-    pub fn decode(&mut self, au: &[u8]) -> Result<Option<PyroWavePlanarFrame>> {
+    /// `wire_window` holds whole self-delimiting packets, zero-padded — walk and strip);
+    /// `complete` = every shard arrived (a partial decodes anyway: missing blocks are
+    /// localized blur for exactly this frame, §4.4).
+    pub fn decode_frame(
+        &mut self,
+        au: &[u8],
+        aligned: bool,
+        complete: bool,
+    ) -> Result<Option<PyroWavePlanarFrame>> {
         // SAFETY: single decode thread; all handles owned/pinned by `self`; queue access
         // serialized under the device-wide QueueLock; the fence bounds GPU completion
         // before the frame is handed to the presenter.
-        unsafe { self.decode_inner(au) }
+        unsafe { self.decode_inner(au, aligned, complete) }
     }
 
-    unsafe fn decode_inner(&mut self, au: &[u8]) -> Result<Option<PyroWavePlanarFrame>> {
+    /// Consume one framed shard window (§4.4): a 4-byte prefix (u16 used-length + u16
-        pw_check(
+    /// kind) then either WHOLE self-delimiting codec packets (PACKED) or one fragment of
-            pw::pyrowave_decoder_push_packet(self.pw_dec, au.as_ptr() as *const c_void, au.len()),
+    /// an oversized packet (FRAG chain). A lost shard arrives as a zeroed window
-            "push_packet",
+    /// (used = 0) — skipped, and it breaks any fragment chain it interrupts (that
-        )?;
+    /// packet's blocks are unusable without their end; dropping them is the §4.4 blur).
-        // The reassembler delivers complete AUs only, so a frame is ready per push; a
+    unsafe fn push_window(&mut self, win: &[u8], frag: &mut Vec<u8>) -> Result<()> {
-        // stale/duplicate packet (sequence rewind) simply isn't — skip, no error.
+        if win.len() < 4 {
-        if !pw::pyrowave_decoder_decode_is_ready(self.pw_dec, false) {
+            return Ok(());
+        }
+        let used = u16::from_le_bytes([win[0], win[1]]) as usize;
+        let kind = u16::from_le_bytes([win[2], win[3]]);
+        if used == 0 || 4 + used > win.len() {
+            frag.clear(); // missing / garbage window — drop any chain in progress
+            return Ok(());
+        }
+        let body = &win[4..4 + used];
+        match kind {
+            0 => {
+                frag.clear();
+                pw_check(
+                    pw::pyrowave_decoder_push_packet(
+                        self.pw_dec,
+                        body.as_ptr() as *const c_void,
+                        body.len(),
+                    ),
+                    "push_packet",
+                )
+            }
+            1 => {
+                frag.clear();
+                frag.extend_from_slice(body);
+                Ok(())
+            }
+            2 => {
+                if !frag.is_empty() {
+                    frag.extend_from_slice(body);
+                }
+                Ok(())
+            }
+            3 => {
+                if !frag.is_empty() {
+                    frag.extend_from_slice(body);
+                    let r = pw_check(
+                        pw::pyrowave_decoder_push_packet(
+                            self.pw_dec,
+                            frag.as_ptr() as *const c_void,
+                            frag.len(),
+                        ),
+                        "push_packet (fragmented)",
+                    );
+                    frag.clear();
+                    return r;
+                }
+                Ok(())
+            }
+            _ => {
+                frag.clear();
+                Ok(())
+            }
+        }
+    }
+
+    unsafe fn decode_inner(
+        &mut self,
+        au: &[u8],
+        aligned: bool,
+        complete: bool,
+    ) -> Result<Option<PyroWavePlanarFrame>> {
+        if aligned {
+            let mut frag: Vec<u8> = Vec::new();
+            for win in au.chunks(self.wire_window) {
+                self.push_window(win, &mut frag)?;
+            }
+        } else {
+            pw_check(
+                pw::pyrowave_decoder_push_packet(
+                    self.pw_dec,
+                    au.as_ptr() as *const c_void,
+                    au.len(),
+                ),
+                "push_packet",
+            )?;
+        }
+        // A complete AU that isn't ready is a stale/duplicate (sequence rewind) — skip.
+        // A PARTIAL is decoded regardless: missing wavelet blocks reconstruct as zeros,
+        // i.e. localized blur for exactly this one frame (the next is complete again).
+        if complete && !pw::pyrowave_decoder_decode_is_ready(self.pw_dec, false) {
             return Ok(None);
         }
 

crates/pf-presenter/src/vk.rs

@@ -1254,7 +1254,7 @@ impl Presenter {
                 .csc_planar
                 .as_ref()
                 .context("PyroWave frame but the device failed the pyrowave probe")?;
-            planar.bind_planes_planar(&self.device, f.views.map(|v| vk::ImageView::from_raw(v)));
+            planar.bind_planes_planar(&self.device, f.views.map(vk::ImageView::from_raw));
         }
         if let Some(o) = overlay {
             // Point the composite at this overlay image (same fence-wait safety).

crates/punktfunk-core/src/client.rs

@@ -71,6 +71,8 @@ enum CtrlRequest {
 #[derive(Clone, Copy)]
 struct Negotiated {
     mode: Mode,
+    /// Wire shard payload — the chunk-aligned parse window (plan §4.4).
+    shard_payload: u16,
     compositor: CompositorPref,
     gamepad: GamepadPref,
     /// SHA-256 of the certificate the host actually presented (TOFU callers persist this).
@@ -513,6 +515,9 @@ pub struct NativeClient {
     /// requested rate clamped to the host's range, or its default if we requested `0`. `0` = an
     /// older host that didn't report it.
     pub resolved_bitrate_kbps: u32,
+    /// The session's wire shard payload (bytes of AU per datagram) — the parse-window size
+    /// for chunk-aligned AUs ([`crate::packet::USER_FLAG_CHUNK_ALIGNED`], plan §4.4).
+    pub shard_payload: u16,
     /// Host clock minus client clock (ns), from the connect-time skew handshake. Add it to a local
     /// receive/present timestamp to express it in the host's capture clock (the AU `pts_ns`), making
     /// glass-to-glass latency valid across machines. `0` = no correction (an old host that didn't
@@ -778,6 +783,7 @@ impl NativeClient {
             resolved_compositor: negotiated.compositor,
             resolved_gamepad: negotiated.gamepad,
             resolved_bitrate_kbps: negotiated.bitrate_kbps,
+            shard_payload: negotiated.shard_payload,
             clock_offset_ns: negotiated.clock_offset_ns,
             bit_depth: negotiated.bit_depth,
             color: negotiated.color,
@@ -1540,7 +1546,14 @@ async fn worker_main(args: WorkerArgs) {
             if let Ok(sock) = transport.try_clone_socket() {
                 crate::transport::spawn_data_punch(sock, shutdown.clone());
             }
-            let session = Session::new(welcome.session_config(Role::Client), Box::new(transport))?;
+            let mut session =
+                Session::new(welcome.session_config(Role::Client), Box::new(transport))?;
+            // PyroWave sessions opt into partial delivery (plan §4.4): an aged-out lossy
+            // frame arrives as blocks-with-holes instead of vanishing — the all-intra codec
+            // renders it as one frame of localized blur, strictly better than a freeze.
+            if welcome.codec == crate::quic::CODEC_PYROWAVE {
+                session.set_deliver_partial_frames(true);
+            }
             Ok::<_, PunktfunkError>((
                 session,
                 send,
@@ -1558,6 +1571,7 @@ async fn worker_main(args: WorkerArgs) {
                     chroma_format: welcome.chroma_format,
                     audio_channels: welcome.audio_channels,
                     codec: welcome.codec,
+                    shard_payload: welcome.shard_payload,
                 },
                 welcome.host_caps,
             ))
@@ -2490,6 +2504,7 @@ mod frame_channel_tests {
             frame_index: i,
             pts_ns: i as u64,
             flags: 0,
+            complete: true,
         }
     }
 

crates/punktfunk-core/src/packet.rs

@@ -54,6 +54,14 @@ pub const USER_FLAG_RECOVERY_POINT: u32 = 0x10;
 /// `AV_FRAME_FLAG_KEY` — this host flag is the only signal.
 pub const USER_FLAG_RECOVERY_ANCHOR: u32 = 0x20;
 
+/// `user_flags` bit: the AU's content is **shard-aligned self-delimiting chunks** — every
+/// `shard_payload`-sized window of the frame buffer starts a fresh codec packet, padded to the
+/// window with zeros (PyroWave datagram-aligned mode, design/pyrowave-codec-plan.md §4.4). Two
+/// consequences: a receiver that opted into partial delivery can use an aged-out frame's buffer
+/// AS-IS (missing shards stay zeroed; the codec's block walk skips zero windows), and even a
+/// COMPLETE frame must be consumed window-by-window (the padding is not part of the stream).
+pub const USER_FLAG_CHUNK_ALIGNED: u32 = 0x40;
+
 /// Widest lost-frame range (frames, wrapping `last - first`) a reference-frame-invalidation
 /// recovery may be asked to repair; anything wider goes straight to the keyframe path on BOTH
 /// ends. RFI can only re-reference history the encoder still holds — NVENC keeps a 5-frame DPB,
@@ -89,6 +97,13 @@ const LOSS_WINDOW_NS: u64 = 120_000_000;
 /// 120 ms ≈ 14 indices, so 64 leaves ample slack up to ~500 fps.
 const HARD_LOSS_WINDOW: u32 = 64;
 
+/// The much tighter fuse for PARTIAL-deliverable frames (chunk-aligned AUs with a consumer
+/// that opted in): once anything newer exists and this much capture time passed, the frame
+/// is delivered as-is — its stragglers can only make it less late, and each frame is
+/// independently decodable, so waiting the full loss window (120 ms) would inject ancient
+/// frames into a live stream. ~2 frame periods at 60 fps rides out normal reorder.
+const PARTIAL_WINDOW_NS: u64 = 30_000_000;
+
 /// How many frames behind the newest the reassembler remembers emitted/abandoned frame indices
 /// (`completed`), so a straggler shard can neither resurrect an abandoned frame nor re-open an
 /// emitted one. Must cover at least [`HARD_LOSS_WINDOW`]: stragglers can trickle in later than the
@@ -451,6 +466,13 @@ const RECOVERY_POOL_MAX: usize = 512;
 /// Client-side only.
 pub struct Reassembler {
     limits: ReassemblerLimits,
+    /// Deliver aged-out incomplete frames whose AUs are [`USER_FLAG_CHUNK_ALIGNED`] instead of
+    /// silently dropping them (client opt-in — the PyroWave decode path): the frame buffer is
+    /// already the right shape (received shards at their final offsets, zeros elsewhere).
+    /// They still count into `frames_dropped` — a partial IS lost data for the loss reports.
+    deliver_partial: bool,
+    /// The newest such partial awaiting pickup (newest-wins: partials are a lossy byproduct).
+    pending_partial: Option<Frame>,
     /// The video stream's window — its aged-out incomplete frames count into `frames_dropped`
     /// (the client's loss-recovery trigger).
     video: ReassemblyWindow,
@@ -472,6 +494,8 @@ impl Reassembler {
     pub fn new(limits: ReassemblerLimits) -> Self {
         Reassembler {
             limits,
+            deliver_partial: false,
+            pending_partial: None,
             video: ReassemblyWindow::default(),
             probe: ReassemblyWindow::default(),
             recovery_pool: Vec::new(),
@@ -479,6 +503,19 @@ impl Reassembler {
         }
     }
 
+    /// Opt into partial delivery of chunk-aligned frames (see [`Reassembler::deliver_partial`]).
+    pub fn set_deliver_partial(&mut self, on: bool) {
+        self.deliver_partial = on;
+        if !on {
+            self.pending_partial = None;
+        }
+    }
+
+    /// Take the newest aged-out partial frame, if one is pending (see `set_deliver_partial`).
+    pub fn take_partial(&mut self) -> Option<Frame> {
+        self.pending_partial.take()
+    }
+
     /// Ingest one (already-decrypted) packet. Returns the access unit when its last
     /// block completes, otherwise `None`.
     pub fn push(
@@ -507,6 +544,8 @@ impl Reassembler {
         // in-flight budget are all touched while a frame entry is mutably borrowed.
         let Reassembler {
             limits,
+            deliver_partial,
+            pending_partial,
             video,
             probe,
             recovery_pool,
@@ -579,6 +618,7 @@ impl Reassembler {
             recovery_pool,
             in_flight_bytes,
             lim.max_data_shards,
+            (*deliver_partial && !is_probe).then_some(pending_partial),
         );
 
         // Drop shards for frames already terminated (emitted — e.g. the recovery shards of a
@@ -756,6 +796,7 @@ impl Reassembler {
                 frame_index: hdr.frame_index,
                 pts_ns: done.pts_ns,
                 flags: done.user_flags,
+                complete: true,
             }));
         }
         Ok(None)
@@ -809,6 +850,8 @@ impl ReassemblyWindow {
         recovery_pool: &mut Vec<Vec<u8>>,
         in_flight_bytes: &mut usize,
         max_data_shards: usize,
+        // `Some(sink)` = deliver aged-out CHUNK_ALIGNED frames instead of only dropping them.
+        mut partial_sink: Option<&mut Option<Frame>>,
     ) {
         let (newest, newest_pts) = match self.newest_frame {
             // `frame_index` is newer iff it's within the forward half of the index space.
@@ -819,9 +862,17 @@ impl ReassemblyWindow {
 
         let before = self.frames.len();
         let completed = &mut self.completed;
+        let partial_on = partial_sink.is_some();
         self.frames.retain(|&idx, f| {
+            // Partial-deliverable frames age out on the TIGHT fuse (see PARTIAL_WINDOW_NS);
+            // everything else keeps the full loss window.
+            let window_ns = if partial_on && f.user_flags & USER_FLAG_CHUNK_ALIGNED != 0 {
+                PARTIAL_WINDOW_NS
+            } else {
+                LOSS_WINDOW_NS
+            };
             let keep = newest.wrapping_sub(idx) <= HARD_LOSS_WINDOW
-                && newest_pts.saturating_sub(f.pts_ns) <= LOSS_WINDOW_NS;
+                && newest_pts.saturating_sub(f.pts_ns) <= window_ns;
             if !keep {
                 // Remember the abandoned index so a straggler shard is dropped (below, and in
                 // `push`) instead of resurrecting the frame — which would re-allocate its buffers
@@ -831,6 +882,28 @@ impl ReassemblyWindow {
                 completed.insert(idx, reconstructed_shards(&f.blocks, max_data_shards));
                 // Release its buffer budget and reclaim its parity bufs for the pool.
                 *in_flight_bytes -= f.buf.len();
+                // Partial delivery (chunk-aligned AUs only): the buffer is already exactly
+                // what the consumer needs — received shards at their final offsets, zeros
+                // where shards are missing (the codec's block walk skips zero windows).
+                // Newest-wins if several age out in one prune. Still counted dropped below.
+                if let Some(sink) = partial_sink.as_deref_mut() {
+                    if f.user_flags & USER_FLAG_CHUNK_ALIGNED != 0 {
+                        let mut buf = std::mem::take(&mut f.buf);
+                        buf.truncate(f.frame_bytes);
+                        let newer = sink
+                            .as_ref()
+                            .is_none_or(|p| idx.wrapping_sub(p.frame_index) <= u32::MAX / 2);
+                        if newer {
+                            *sink = Some(Frame {
+                                data: buf,
+                                frame_index: idx,
+                                pts_ns: f.pts_ns,
+                                flags: f.user_flags,
+                                complete: false,
+                            });
+                        }
+                    }
+                }
                 for block in f.blocks.values_mut() {
                     for slot in block.recovery.iter_mut() {
                         if let Some(rb) = slot.take() {

crates/punktfunk-core/src/session.rs

@@ -24,6 +24,12 @@ pub struct Frame {
     pub frame_index: u32,
     pub pts_ns: u64,
     pub flags: u32,
+    /// `false` = a partial delivery: the frame aged out of the loss window with shards
+    /// missing, and the session opted into receiving it anyway
+    /// ([`Session::set_deliver_partial_frames`]). Only chunk-aligned AUs
+    /// ([`crate::packet::USER_FLAG_CHUNK_ALIGNED`]) are ever delivered partial; missing
+    /// shard ranges are zero-filled at their exact offsets.
+    pub complete: bool,
 }
 
 /// One end of a stream. Constructed for a single [`Role`]; calling the other role's
@@ -634,6 +640,20 @@ impl Session {
 
     /// Client: drain the transport until a whole access unit is recovered, or no more
     /// packets are pending ([`PunktfunkError::NoFrame`]).
+    /// Client opt-in: deliver aged-out incomplete chunk-aligned frames as
+    /// [`Frame`]`{ complete: false }` instead of only dropping them (the PyroWave
+    /// datagram-aligned mode, plan §4.4 — a lost datagram costs a few blocks of blur,
+    /// not the frame). No effect on other codecs' AUs (they never carry the flag).
+    pub fn set_deliver_partial_frames(&mut self, on: bool) {
+        self.reassembler.set_deliver_partial(on);
+    }
+
+    /// The session's negotiated wire shard payload size (bytes of AU per datagram) —
+    /// the window size for chunk-aligned AUs (`USER_FLAG_CHUNK_ALIGNED`).
+    pub fn shard_payload(&self) -> usize {
+        self.config.shard_payload
+    }
+
     pub fn poll_frame(&mut self) -> Result<Frame> {
         if self.config.role != Role::Client {
             return Err(PunktfunkError::InvalidArg(
@@ -661,6 +681,11 @@ impl Session {
                 }
                 self.recv_idx = 0;
                 if self.recv_count == 0 {
+                    // Nothing new on the wire — hand over an aged-out partial if one is
+                    // waiting (it can only get staler).
+                    if let Some(p) = self.reassembler.take_partial() {
+                        return Ok(p);
+                    }
                     return Err(PunktfunkError::NoFrame);
                 }
             }
@@ -725,6 +750,11 @@ impl Session {
                 StatsCounters::add(&self.stats.frames_completed, 1);
                 return Ok(frame);
             }
+            // A push that completed nothing may still have aged a partial out — deliver it
+            // ahead of further draining (its successors are already arriving).
+            if let Some(p) = self.reassembler.take_partial() {
+                return Ok(p);
+            }
         }
     }
 
@@ -980,6 +1010,101 @@ mod wire_equivalence_tests {
     fn pattern(len: usize) -> Vec<u8> {
         (0..len).map(|i| (i * 31 + 7) as u8).collect()
     }
+
+    /// Partial delivery (plan §4.4): a chunk-aligned frame that loses shards past FEC's
+    /// reach is DELIVERED once it ages out — `complete: false`, received shards at their
+    /// exact offsets, missing ranges zero-filled — instead of silently dropping. Plain
+    /// AUs (no flag) keep the drop behavior even with the opt-in enabled.
+    #[test]
+    fn partial_delivery_of_chunk_aligned_frames() {
+        use crate::packet::USER_FLAG_CHUNK_ALIGNED;
+        let mk = |role| Config {
+            role,
+            phase: ProtocolPhase::P2Punktfunk,
+            fec: FecConfig {
+                scheme: FecScheme::Gf16,
+                fec_percent: 0, // no parity — any drop leaves a hole
+                max_data_per_block: 64,
+            },
+            shard_payload: 1024,
+            max_frame_bytes: 8 * 1024 * 1024,
+            encrypt: false,
+            key: [0u8; 16],
+            salt: [0u8; 4],
+            loopback_drop_period: 0,
+        };
+        // Drop exactly one datagram (the 3rd) out of the first frame's shards.
+        let (h, c) = crate::transport::loopback_pair(3, 1);
+        let mut host = Session::new(mk(Role::Host), Box::new(h)).unwrap();
+        let mut client = Session::new(mk(Role::Client), Box::new(c)).unwrap();
+        client.set_deliver_partial_frames(true);
+
+        // 8 shards of chunk-aligned payload.
+        let frame = pattern(8 * 1024);
+        host.submit_frame(&frame, 1_000, USER_FLAG_CHUNK_ALIGNED)
+            .unwrap();
+        // The incomplete frame ages out on the HARD index window — push enough newer
+        // (complete) frames past it. Collect everything the client emits.
+        let mut got_partial = None;
+        let mut completes = 0;
+        for i in 0..80u64 {
+            let filler = pattern(1024);
+            host.submit_frame(&filler, 2_000 + i, USER_FLAG_CHUNK_ALIGNED)
+                .unwrap();
+            loop {
+                match client.poll_frame() {
+                    Ok(f) if !f.complete => got_partial = Some(f),
+                    Ok(_) => completes += 1,
+                    Err(PunktfunkError::NoFrame) => break,
+                    Err(e) => panic!("unexpected: {e}"),
+                }
+            }
+        }
+        let p = got_partial.expect("the lossy frame must be delivered partial");
+        assert_eq!(p.pts_ns, 1_000);
+        assert_eq!(p.data.len(), frame.len());
+        assert!(p.flags & USER_FLAG_CHUNK_ALIGNED != 0);
+        // Exactly one 1024-byte shard is zeroed; every other offset matches the original.
+        let mut zero_windows = 0;
+        for w in 0..8 {
+            let win = &p.data[w * 1024..(w + 1) * 1024];
+            if win.iter().all(|&b| b == 0) {
+                zero_windows += 1;
+            } else {
+                assert_eq!(win, &frame[w * 1024..(w + 1) * 1024], "window {w} corrupt");
+            }
+        }
+        // loopback_pair(3, _) drops every 3rd datagram, so several of the 8 shards are
+        // gone — the exact count depends on phase; what matters is that SOME are zeroed
+        // and every survivor is intact.
+        assert!(
+            (1..8).contains(&zero_windows),
+            "dropped shards zero-filled (got {zero_windows})"
+        );
+        assert!(completes > 40, "surviving filler frames flow normally");
+
+        // Control: WITHOUT the flag the same loss is a plain drop, opt-in or not.
+        let (h2, c2) = crate::transport::loopback_pair(3, 1);
+        let mut host2 = Session::new(mk(Role::Host), Box::new(h2)).unwrap();
+        let mut client2 = Session::new(mk(Role::Client), Box::new(c2)).unwrap();
+        client2.set_deliver_partial_frames(true);
+        host2.submit_frame(&pattern(8 * 1024), 1_000, 0).unwrap();
+        let mut saw_partial = false;
+        for i in 0..80u64 {
+            host2.submit_frame(&pattern(1024), 2_000 + i, 0).unwrap();
+            loop {
+                match client2.poll_frame() {
+                    Ok(f) => saw_partial |= !f.complete,
+                    Err(PunktfunkError::NoFrame) => break,
+                    Err(e) => panic!("unexpected: {e}"),
+                }
+            }
+        }
+        assert!(
+            !saw_partial,
+            "unflagged AUs must never be delivered partial"
+        );
+    }
 }
 
 #[cfg(test)]

crates/punktfunk-host/src/encode.rs

@@ -29,6 +29,10 @@ pub struct EncodedFrame {
     /// clean re-anchor). Without it the client's freeze can only lift on an IDR — which the host
     /// suppresses after a successful RFI (the cooldown), a ~1 s frozen stall per loss event.
     pub recovery_anchor: bool,
+    /// The AU is shard-aligned self-delimiting chunks (see [`Encoder::set_wire_chunking`]);
+    /// the session stamps [`punktfunk_core::packet::USER_FLAG_CHUNK_ALIGNED`] so the client
+    /// windows its parse and may opt into partial delivery. Only the PyroWave backend sets it.
+    pub chunk_aligned: bool,
 }
 
 /// Codec selection negotiated with the client.
@@ -356,6 +360,13 @@ pub trait Encoder: Send {
     fn reconfigure_bitrate(&mut self, _bps: u64) -> bool {
         false
     }
+    /// Wire-chunk the encoder's AUs at the session's shard payload size (the PyroWave
+    /// datagram-aligned mode, plan §4.4): every `shard_payload` window of the emitted AU
+    /// starts a fresh self-delimiting codec packet, zero-padded to the window — so a lost
+    /// datagram costs a few coefficient blocks, not the frame. AUs produced this way are
+    /// flagged [`EncodedFrame::chunk_aligned`] and the session marks them on the wire.
+    /// Default: no-op (the H.26x backends' bitstreams cannot be cut losslessly).
+    fn set_wire_chunking(&mut self, _shard_payload: usize) {}
     /// Signal end-of-stream. After this, drain the remaining AUs with [`poll`](Self::poll)
     /// until it returns `None` — NVENC buffers frames internally even at `delay=0`.
     fn flush(&mut self) -> Result<()>;
@@ -519,6 +530,12 @@ impl Encoder for TrackedEncoder {
     fn invalidate_ref_frames(&mut self, first_frame: i64, last_frame: i64) -> bool {
         self.inner.invalidate_ref_frames(first_frame, last_frame)
     }
+    // The classic TrackedEncoder trap: a defaulted trait method that isn't forwarded
+    // silently no-ops through the wrapper (bit the direct-NVENC work, then THIS — the
+    // §4.4 chunking probe run hit the default while the plan said Some(1408)).
+    fn set_wire_chunking(&mut self, shard_payload: usize) {
+        self.inner.set_wire_chunking(shard_payload)
+    }
     fn poll(&mut self) -> Result<Option<EncodedFrame>> {
         self.inner.poll()
     }

crates/punktfunk-host/src/encode/linux/mod.rs

@@ -652,6 +652,7 @@ impl Encoder for NvencEncoder {
                     pts_ns,
                     keyframe: pkt.is_key(),
                     recovery_anchor: false,
+                    chunk_aligned: false,
                 }))
             }
             // No packet ready yet (need another input frame).

crates/punktfunk-host/src/encode/linux/nvenc_cuda.rs

@@ -1148,6 +1148,7 @@ impl Encoder for NvencCudaEncoder {
                 pts_ns,
                 keyframe,
                 recovery_anchor: anchor,
+                chunk_aligned: false,
             }))
         }
     }

crates/punktfunk-host/src/encode/linux/pyrowave.rs

@@ -43,6 +43,13 @@ const IMPORT_CACHE_CAP: usize = 16;
 /// Headroom over the per-frame rate budget for the packetized bitstream (block headers + meta;
 /// the rate controller itself never exceeds the budget).
 const BS_SLACK: usize = 256 * 1024;
+/// Chunked-mode window framing (§4.4): 4-byte prefix per shard-sized window.
+const WINDOW_PREFIX: usize = 4;
+/// Window kinds: whole packets / an oversized packet's fragments.
+const WIN_PACKED: u16 = 0;
+const WIN_FRAG_FIRST: u16 = 1;
+const WIN_FRAG_CONT: u16 = 2;
+const WIN_FRAG_LAST: u16 = 3;
 
 /// The DRM modifiers the PyroWave device can import as a SAMPLED image of the capture's
 /// packed-RGB format. The capture advertises these for the pyrowave passthrough instead of
@@ -178,6 +185,10 @@ pub struct PyroWaveEncoder {
     fps: u32,
     /// Per-frame bitstream budget (hard CBR): `bitrate / (8 * fps)`.
     frame_budget: usize,
+    /// Datagram-aligned mode (plan §4.4): packetize at this boundary and pad every codec
+    /// packet to it, so each wire shard carries whole self-delimiting packets. `None` =
+    /// one packet per AU (the dense MVP shape).
+    wire_chunk: Option<usize>,
     bitstream: Vec<u8>,
     pending: VecDeque<EncodedFrame>,
     frame_count: u64,
@@ -532,6 +543,7 @@ impl PyroWaveEncoder {
             height: h,
             fps,
             frame_budget,
+            wire_chunk: None,
             bitstream: Vec::new(),
             pending: VecDeque::new(),
             frame_count: 0,
@@ -865,31 +877,100 @@ impl PyroWaveEncoder {
         dev.wait_for_fences(&[self.fence], true, 5_000_000_000)
             .context("pyrowave encode fence")?;
 
-        // ---- packetize: boundary = whole buffer, so the AU is exactly one pyrowave packet ----
+        // ---- packetize ----
+        // Dense (default): boundary = whole buffer → the AU is exactly one pyrowave packet.
+        // Datagram-aligned (§4.4, `set_wire_chunking`): boundary = the wire shard payload;
+        // each codec packet is zero-padded to the boundary so every shard carries whole
+        // self-delimiting packets — the client windows its parse and a lost shard costs
+        // only those blocks. Padding cost is small: the packetizer fills close to the
+        // boundary by design.
         let cap = self.frame_budget + BS_SLACK;
         self.bitstream.resize(cap, 0);
+        // Chunked mode reserves 4 bytes per window for the framing prefix.
+        let boundary = self.wire_chunk.map(|c| c - WINDOW_PREFIX).unwrap_or(cap);
         let mut n: usize = 0;
         pw_check(
-            pw::pyrowave_encoder_compute_num_packets(self.pw_enc, cap, &mut n),
+            pw::pyrowave_encoder_compute_num_packets(self.pw_enc, boundary, &mut n),
             "compute_num_packets",
         )?;
-        if n != 1 {
+        if n == 0 || (self.wire_chunk.is_none() && n != 1) {
-            bail!("pyrowave: expected a single packet at boundary {cap}, got {n}");
+            bail!("pyrowave: unexpected packet count {n} at boundary {boundary}");
         }
-        let mut packet = pw::pyrowave_packet { offset: 0, size: 0 };
+        let mut packets = vec![pw::pyrowave_packet { offset: 0, size: 0 }; n];
         let mut out_n: usize = 0;
         pw_check(
             pw::pyrowave_encoder_packetize(
                 self.pw_enc,
-                &mut packet,
+                packets.as_mut_ptr(),
-                cap,
+                boundary,
                 &mut out_n,
                 self.bitstream.as_mut_ptr() as *mut std::ffi::c_void,
                 cap,
             ),
             "packetize",
         )?;
-        let au = self.bitstream[packet.offset..packet.offset + packet.size].to_vec();
+        packets.truncate(out_n.max(1));
+        let au = if let Some(chunk) = self.wire_chunk {
+            // Window framing (§4.4): each `chunk`-sized window opens with a 4-byte prefix
+            // (u16 used-length + u16 kind) and carries either WHOLE self-delimiting codec
+            // packets (PACKED — several small ones share a window) or one fragment of an
+            // oversized packet (FRAG chain — pyrowave 32×32 blocks are atomic and may
+            // exceed a shard). A lost shard zeroes its window (used = 0) — the receiver
+            // skips it and drops any fragment chain it interrupts.
+            let payload_max = chunk - WINDOW_PREFIX;
+            let mut au: Vec<u8> = Vec::with_capacity((packets.len() + 1) * chunk);
+            // The currently-open PACKED window: (start offset of its prefix, bytes used).
+            let mut open: Option<(usize, usize)> = None;
+            let close = |au: &mut Vec<u8>, open: &mut Option<(usize, usize)>, chunk: usize| {
+                if let Some((start, used)) = open.take() {
+                    au[start..start + 2].copy_from_slice(&(used as u16).to_le_bytes());
+                    au[start + 2..start + 4].copy_from_slice(&WIN_PACKED.to_le_bytes());
+                    au.resize(start + chunk, 0);
+                }
+            };
+            for p in &packets {
+                let bytes = &self.bitstream[p.offset..p.offset + p.size];
+                if p.size <= payload_max {
+                    let fits = open.is_some_and(|(_, used)| used + p.size <= payload_max);
+                    if !fits {
+                        close(&mut au, &mut open, chunk);
+                        let start = au.len();
+                        au.resize(start + WINDOW_PREFIX, 0);
+                        open = Some((start, 0));
+                    }
+                    au.extend_from_slice(bytes);
+                    if let Some((_, used)) = open.as_mut() {
+                        *used += p.size;
+                    }
+                } else {
+                    // Oversized packet: its own FRAG chain of full windows.
+                    close(&mut au, &mut open, chunk);
+                    let mut off = 0usize;
+                    while off < p.size {
+                        let take = (p.size - off).min(payload_max);
+                        let kind = if off == 0 {
+                            WIN_FRAG_FIRST
+                        } else if off + take == p.size {
+                            WIN_FRAG_LAST
+                        } else {
+                            WIN_FRAG_CONT
+                        };
+                        let start = au.len();
+                        au.resize(start + WINDOW_PREFIX, 0);
+                        au[start..start + 2].copy_from_slice(&(take as u16).to_le_bytes());
+                        au[start + 2..start + 4].copy_from_slice(&kind.to_le_bytes());
+                        au.extend_from_slice(&bytes[off..off + take]);
+                        au.resize(start + chunk, 0);
+                        off += take;
+                    }
+                }
+            }
+            close(&mut au, &mut open, chunk);
+            au
+        } else {
+            let p = &packets[0];
+            self.bitstream[p.offset..p.offset + p.size].to_vec()
+        };
         self.frame_count += 1;
         self.pending.push_back(EncodedFrame {
             data: au,
@@ -898,6 +979,7 @@ impl PyroWaveEncoder {
             // whole recovery story (plan §1.2).
             keyframe: true,
             recovery_anchor: false,
+            chunk_aligned: self.wire_chunk.is_some(),
         });
         Ok(())
     }
@@ -960,6 +1042,17 @@ impl Encoder for PyroWaveEncoder {
         true
     }
 
+    fn set_wire_chunking(&mut self, shard_payload: usize) {
+        // Sanity floor: a boundary below one block header + payload word is meaningless.
+        if shard_payload >= 64 {
+            self.wire_chunk = Some(shard_payload);
+            tracing::info!(
+                shard_payload,
+                "pyrowave: datagram-aligned packetization on (partial-frame loss mode)"
+            );
+        }
+    }
+
     fn flush(&mut self) -> Result<()> {
         // Synchronous per-frame encode: nothing buffered beyond `pending`.
         Ok(())
@@ -1127,6 +1220,83 @@ mod tests {
             );
         }
 
+        // Datagram-aligned mode (§4.4): every emitted AU is a whole number of framed
+        // windows — 4-byte prefix (used-length + kind), whole packets or FRAG chains for
+        // oversized atomic blocks, zero padding after `used`. Walking + reassembling the
+        // fragments must reproduce a decodable packet stream.
+        enc.set_wire_chunking(1408);
+        enc.submit(&cpu_frame(w, h, 500, [90, 60, 30, 255]))
+            .expect("chunked submit");
+        let au = enc.poll().expect("poll").expect("chunked AU");
+        assert!(au.chunk_aligned);
+        assert_eq!(au.data.len() % 1408, 0, "AU is a whole number of windows");
+        // SAFETY: test-only FFI with locally-owned buffers.
+        unsafe {
+            let mut dev: pw::pyrowave_device = std::ptr::null_mut();
+            assert_eq!(
+                pw::pyrowave_create_default_device(&mut dev),
+                pw::pyrowave_result_PYROWAVE_SUCCESS
+            );
+            let dinfo = pw::pyrowave_decoder_create_info {
+                device: dev,
+                width: w as i32,
+                height: h as i32,
+                chroma: pw::pyrowave_chroma_subsampling_PYROWAVE_CHROMA_SUBSAMPLING_420,
+                fragment_path: false,
+            };
+            let mut dec: pw::pyrowave_decoder = std::ptr::null_mut();
+            assert_eq!(
+                pw::pyrowave_decoder_create(&dinfo, &mut dec),
+                pw::pyrowave_result_PYROWAVE_SUCCESS
+            );
+            let mut frag: Vec<u8> = Vec::new();
+            let mut pushed = 0usize;
+            for win in au.data.chunks(1408) {
+                let used = u16::from_le_bytes([win[0], win[1]]) as usize;
+                let kind = u16::from_le_bytes([win[2], win[3]]);
+                assert!(4 + used <= win.len(), "window overrun");
+                assert!(win[4 + used..].iter().all(|&b| b == 0), "non-zero padding");
+                let body = &win[4..4 + used];
+                match kind {
+                    0 => {
+                        assert_eq!(
+                            pw::pyrowave_decoder_push_packet(
+                                dec,
+                                body.as_ptr() as *const _,
+                                body.len()
+                            ),
+                            pw::pyrowave_result_PYROWAVE_SUCCESS
+                        );
+                        pushed += body.len();
+                    }
+                    1 => frag = body.to_vec(),
+                    2 => frag.extend_from_slice(body),
+                    3 => {
+                        frag.extend_from_slice(body);
+                        assert_eq!(
+                            pw::pyrowave_decoder_push_packet(
+                                dec,
+                                frag.as_ptr() as *const _,
+                                frag.len()
+                            ),
+                            pw::pyrowave_result_PYROWAVE_SUCCESS
+                        );
+                        pushed += frag.len();
+                        frag.clear();
+                    }
+                    k => panic!("unknown window kind {k}"),
+                }
+            }
+            assert!(pushed > 0, "chunked AU carries real packets");
+            assert!(
+                pw::pyrowave_decoder_decode_is_ready(dec, false),
+                "chunked AU incomplete after framed walk"
+            );
+            pw::pyrowave_decoder_destroy(dec);
+            pw::pyrowave_device_destroy(dev);
+        }
+        enc.set_wire_chunking(0); // below the floor — back to dense
+
         // In-place rate retarget + encoder rebuild both keep encoding.
         assert!(enc.reconfigure_bitrate(100_000_000));
         assert!(enc.reset());

crates/punktfunk-host/src/encode/linux/vaapi.rs

@@ -297,6 +297,7 @@ fn poll_encoder(enc: &mut encoder::video::Encoder, fps: u32) -> Result<Option<En
                 pts_ns: pts * 1_000_000_000 / fps as u64,
                 keyframe: pkt.is_key(),
                 recovery_anchor: false,
+                chunk_aligned: false,
             }))
         }
         Err(ffmpeg::Error::Other { errno })

crates/punktfunk-host/src/encode/linux/vulkan_video.rs

@@ -1713,6 +1713,7 @@ impl VulkanVideoEncoder {
             pts_ns: f.pts_ns,
             keyframe: f.keyframe,
             recovery_anchor: f.recovery_anchor,
+            chunk_aligned: false,
         })
     }
 

crates/punktfunk-host/src/encode/sw.rs

@@ -216,6 +216,7 @@ impl Encoder for OpenH264Encoder {
                 pts_ns,
                 keyframe,
                 recovery_anchor: false,
+                chunk_aligned: false,
             });
         }
         self.frame_idx += 1;

crates/punktfunk-host/src/encode/windows/amf.rs

@@ -1937,6 +1937,7 @@ unsafe fn drain_one_output(
         pts_ns,
         keyframe: key_prop || forced,
         recovery_anchor,
+        chunk_aligned: false,
     }))
 }
 

crates/punktfunk-host/src/encode/windows/ffmpeg_win.rs

@@ -340,6 +340,7 @@ fn poll_encoder(enc: &mut encoder::video::Encoder, fps: u32) -> Result<PollOutco
                 pts_ns: pts * 1_000_000_000 / fps as u64,
                 keyframe: pkt.is_key(),
                 recovery_anchor: false,
+                chunk_aligned: false,
             }))
         }
         Err(ffmpeg::Error::Other { errno })

crates/punktfunk-host/src/encode/windows/nvenc.rs

@@ -1224,6 +1224,7 @@ impl NvencD3d11Encoder {
                 pts_ns,
                 keyframe,
                 recovery_anchor: anchor,
+                chunk_aligned: false,
             });
         Ok(())
     }
@@ -1649,6 +1650,7 @@ impl Encoder for NvencD3d11Encoder {
                 pts_ns,
                 keyframe,
                 recovery_anchor: anchor,
+                chunk_aligned: false,
             }))
         }
     }

crates/punktfunk-host/src/pipeline.rs

@@ -27,6 +27,7 @@ pub fn pump_once(
         pts_ns,
         keyframe,
         recovery_anchor,
+        chunk_aligned: _,
     }) = encoder.poll()?
     {
         let mut flags = FLAG_PIC as u32;

crates/punktfunk-host/src/punktfunk1.rs

@@ -3978,7 +3978,12 @@ fn virtual_stream(ctx: SessionContext) -> Result<()> {
     // path now reads this typed `SessionPlan` instead of re-deriving from config at each dispatch site
     // (the latent "capture and encode disagree on the backend" hazard, plan §2.4). `bit_depth` is the
     // only per-session input — capture/topology/encoder are otherwise pure functions of `HostConfig`.
-    let plan = crate::session_plan::SessionPlan::resolve(ctx.bit_depth, ctx.chroma, ctx.codec);
+    let mut plan = crate::session_plan::SessionPlan::resolve(ctx.bit_depth, ctx.chroma, ctx.codec);
+    // PyroWave rides the datagram-aligned wire mode (§4.4): every encoder this session opens
+    // packetizes at the negotiated shard payload, so a lost datagram costs blocks, not frames.
+    if ctx.codec == crate::encode::Codec::PyroWave {
+        plan.wire_chunk = Some(ctx.session.shard_payload());
+    }
     tracing::info!(?plan, "resolved session plan");
     // Single-process path: unpack the context into the locals the loop below uses (names unchanged, so the
     // body is byte-for-byte the same; the receivers are now owned but `try_recv()` is identical).
@@ -4430,12 +4435,15 @@ fn virtual_stream(ctx: SessionContext) -> Result<()> {
                     bit_depth,
                     plan.chroma,
                 ) {
-                    Ok(new_enc) => {
+                    Ok(mut new_enc) => {
                         tracing::info!(
                             from_kbps = bitrate_kbps,
                             to_kbps = new_kbps,
                             "encoder rebuilt at new bitrate (adaptive bitrate)"
                         );
+                        if let Some(c) = plan.wire_chunk {
+                            new_enc.set_wire_chunking(c);
+                        }
                         enc = new_enc;
                         bitrate_kbps = new_kbps;
                         live_bitrate.store(new_kbps, Ordering::Relaxed);
@@ -4866,6 +4874,11 @@ fn virtual_stream(ctx: SessionContext) -> Result<()> {
             if au.recovery_anchor {
                 flags |= punktfunk_core::packet::USER_FLAG_RECOVERY_ANCHOR;
             }
+            // Datagram-aligned PyroWave AU (plan §4.4): the client windows its parse at the
+            // shard payload and may opt into partial delivery of lossy frames.
+            if au.chunk_aligned {
+                flags |= punktfunk_core::packet::USER_FLAG_CHUNK_ALIGNED;
+            }
             // Re-send the HDR mastering metadata (0xCE) on each keyframe (a decoder-resync point) and
             // whenever it changed, so a client that dropped the best-effort datagram re-converges.
             if let Some(m) = last_hdr_meta {
@@ -5194,7 +5207,7 @@ fn build_pipeline(
     };
     // `bit_depth` is the handshake-negotiated value (8, or 10 = HEVC Main10 when the client
     // advertised VIDEO_CAP_10BIT and the host opted in). Threaded down from the Welcome.
-    let enc = crate::encode::open_video(
+    let mut enc = crate::encode::open_video(
         plan.codec,
         frame.format,
         frame.width,
@@ -5206,6 +5219,9 @@ fn build_pipeline(
         plan.chroma,
     )
     .context("open video encoder")?;
+    if let Some(c) = plan.wire_chunk {
+        enc.set_wire_chunking(c);
+    }
     // Post-open cross-check: the Welcome already committed `chroma_format` from the pre-open probe, so
     // warn loudly if the encoder actually opened a different chroma than negotiated (the in-band SPS is
     // authoritative for the decoder, but a mismatch means the probe and the live open disagreed).

crates/punktfunk-host/src/session_plan.rs

@@ -97,6 +97,10 @@ pub struct SessionPlan {
     /// Handshake-negotiated video codec the encoder emits — HEVC by default, H.264 for a GPU-less
     /// software host (`resolve_codec` over the client's advertised codecs ∩ the host's capability).
     pub codec: crate::encode::Codec,
+    /// Datagram-aligned wire chunking for the encoder (plan §4.4): `Some(shard_payload)` on a
+    /// PyroWave session — applied to EVERY encoder this plan opens (initial + all rebuilds) so
+    /// AUs stay shard-aligned across mode/bitrate/stall rebuilds. `None` for the H.26x codecs.
+    pub wire_chunk: Option<usize>,
 }
 
 impl SessionPlan {
@@ -115,6 +119,7 @@ impl SessionPlan {
             hdr: bit_depth >= 10,
             chroma,
             codec,
+            wire_chunk: None,
         }
     }
 

include/punktfunk_core.h

@@ -295,6 +295,14 @@
 // `AV_FRAME_FLAG_KEY` — this host flag is the only signal.
 #define USER_FLAG_RECOVERY_ANCHOR 32
 
+// `user_flags` bit: the AU's content is **shard-aligned self-delimiting chunks** — every
+// `shard_payload`-sized window of the frame buffer starts a fresh codec packet, padded to the
+// window with zeros (PyroWave datagram-aligned mode, design/pyrowave-codec-plan.md §4.4). Two
+// consequences: a receiver that opted into partial delivery can use an aged-out frame's buffer
+// AS-IS (missing shards stay zeroed; the codec's block walk skips zero windows), and even a
+// COMPLETE frame must be consumed window-by-window (the padding is not part of the stream).
+#define USER_FLAG_CHUNK_ALIGNED 64
+
 // Widest lost-frame range (frames, wrapping `last - first`) a reference-frame-invalidation
 // recovery may be asked to repair; anything wider goes straight to the keyframe path on BOTH
 // ends. RFI can only re-reference history the encoder still holds — NVENC keeps a 5-frame DPB,