diff --git a/diagnostics/index.html b/diagnostics/index.html
index 403e39b2d..129965653 100644
--- a/diagnostics/index.html
+++ b/diagnostics/index.html
@@ -485,8 +485,13 @@ <h2 id="detail-title"></h2>
       ws.addEventListener('message', event => {
         try {
           const data = JSON.parse(event.data);
-          if (Array.isArray(data)) {
-            applyBatch(data);
+          // Wire format is one Frame envelope per WebSocket message:
+          //   { type: "Frame", ts_us: <u64>, updates: [...] }
+          // The server only emits a Frame once the capture frontier has
+          // advanced past `ts_us`, so each Frame is a complete view at that
+          // closed logical timestamp.
+          if (data && data.type === 'Frame' && Array.isArray(data.updates)) {
+            applyBatch(data.updates);
           }
         } catch (e) {
           // ignore parse errors
diff --git a/diagnostics/src/server.rs b/diagnostics/src/server.rs
index a0a40c2b6..be7d0ed16 100644
--- a/diagnostics/src/server.rs
+++ b/diagnostics/src/server.rs
@@ -14,7 +14,7 @@
 //! (e.g., `python3 -m http.server 8000`). A future improvement could embed
 //! static file serving here so only one port is needed.
 
-use std::collections::HashMap;
+use std::collections::{BTreeMap, HashMap};
 use std::net::TcpListener;
 use std::sync::mpsc;
 use std::thread;
@@ -155,6 +155,23 @@ fn run_server(port: u16, sink: SinkHandle) {
     let mut clients: HashMap<usize, tungstenite::WebSocket<std::net::TcpStream>> = HashMap::new();
     let mut next_client_id: usize = 0;
 
+    // Per-client buffer of records awaiting their timestamp to close, keyed
+    // by the inner `ts` of each record. A timestamp is "closed" when the
+    // capture stream's frontier (tracked in `progress_counts`) advances past
+    // it — at which point we flush the bucket as a single Frame to the client.
+    let mut pending: HashMap<usize, BTreeMap<Duration, Vec<serde_json::Value>>> =
+        HashMap::new();
+    // Running multiplicities from `Event::Progress` updates. The current
+    // frontier is the smallest key with positive count; an entry at zero is
+    // removed. Anything strictly less than the smallest live key is closed.
+    //
+    // Timely's capture protocol sends progress as *deltas* relative to an
+    // assumed initial frontier of `{T::default(): 1}` — so we must seed the
+    // counter that way, otherwise the first event's `(0ns, -1)` retraction
+    // leaves us at `{0ns: -1}` and the frontier sticks at 0 forever.
+    let mut progress_counts: BTreeMap<Duration, i64> = BTreeMap::new();
+    progress_counts.insert(Duration::default(), 1);
+
     loop {
         // Accept pending connections.
         loop {
@@ -183,17 +200,34 @@ fn run_server(port: u16, sink: SinkHandle) {
             }
         }
 
-        // Drain diagnostic updates and group by client.
-        let mut batches_by_client: HashMap<usize, Vec<serde_json::Value>> = HashMap::new();
+        // Drain diagnostic updates: bucket records by their inner timestamp
+        // per client; absorb progress updates into the running frontier.
+        let mut frontier_changed = false;
         loop {
             match receiver.try_recv() {
-                Ok(Event::Messages(_time, data)) => {
-                    for ((client_id, update), _ts, diff) in data {
+                Ok(Event::Messages(_envelope_time, data)) => {
+                    // The capture envelope time is incidental; the meaningful
+                    // logical time is the per-record `ts`.
+                    for ((client_id, update), ts, diff) in data {
                         let json = update_to_json(&update, diff);
-                        batches_by_client.entry(client_id).or_default().push(json);
+                        pending
+                            .entry(client_id)
+                            .or_default()
+                            .entry(ts)
+                            .or_default()
+                            .push(json);
                     }
                 }
-                Ok(Event::Progress(_)) => {}
+                Ok(Event::Progress(updates)) => {
+                    for (t, diff) in updates {
+                        let entry = progress_counts.entry(t).or_insert(0);
+                        *entry += diff;
+                        if *entry == 0 {
+                            progress_counts.remove(&t);
+                        }
+                    }
+                    frontier_changed = true;
+                }
                 Err(mpsc::TryRecvError::Empty) => break,
                 Err(mpsc::TryRecvError::Disconnected) => {
                     eprintln!("Diagnostics output channel closed, shutting down server");
@@ -206,14 +240,37 @@ fn run_server(port: u16, sink: SinkHandle) {
             }
         }
 
-        // Send batched updates to each client.
+        // If the frontier moved, flush every closed timestamp bucket. One
+        // Frame per closed `ts`, in timestamp order, so each Frame is one
+        // atomic transaction on the client.
         let mut disconnected = Vec::new();
-        for (client_id, updates) in &batches_by_client {
-            if let Some(ws) = clients.get_mut(client_id) {
-                if !updates.is_empty() {
-                    let payload = serde_json::to_string(updates).unwrap();
+        if frontier_changed {
+            // Anything strictly less than the smallest live progress count
+            // is closed. If `progress_counts` is empty, every buffered
+            // timestamp is closed.
+            let frontier: Option<Duration> = progress_counts.keys().next().copied();
+            for (client_id, buckets) in pending.iter_mut() {
+                let closed: Vec<Duration> = match frontier {
+                    Some(f) => buckets.range(..f).map(|(t, _)| *t).collect(),
+                    None => buckets.keys().copied().collect(),
+                };
+                for ts in closed {
+                    let updates = buckets.remove(&ts).unwrap_or_default();
+                    if updates.is_empty() {
+                        continue;
+                    }
+                    let Some(ws) = clients.get_mut(client_id) else {
+                        continue;
+                    };
+                    let frame = serde_json::json!({
+                        "type": "Frame",
+                        "ts_us": ts.as_micros() as u64,
+                        "updates": updates,
+                    });
+                    let payload = serde_json::to_string(&frame).unwrap();
                     if ws.send(Message::Text(payload.into())).is_err() {
                         disconnected.push(*client_id);
+                        break;
                     }
                 }
             }
@@ -229,6 +286,7 @@ fn run_server(port: u16, sink: SinkHandle) {
         }
         for client_id in disconnected {
             clients.remove(&client_id);
+            pending.remove(&client_id);
             client_input.disconnect(client_id, start.elapsed());
             eprintln!("Diagnostics client {client_id} disconnected");
         }