diff --git a/http.go b/http.go
index cae9927..e5b9604 100644
--- a/http.go
+++ b/http.go
@@ -34,6 +34,15 @@ import (
 	"go.mitchellh.com/tack/internal/buildkite"
 )
 
+// wsWriteWait bounds how long any single websocket write (frame or
+// control) is allowed to block before we treat the peer as dead. A
+// client that stops reading but keeps the TCP connection open would
+// otherwise hang the handler indefinitely on a full kernel send buffer.
+// 10s is intentionally generous: real backpressure resolves in
+// milliseconds, so anything past that is a stuck peer we'd rather drop
+// than serve.
+const wsWriteWait = 10 * time.Second
+
 // runHTTP starts the spindle's HTTP server and blocks until ctx is
 // cancelled or the listener returns a fatal error. On ctx cancellation it
 // performs a graceful shutdown with a bounded timeout.
@@ -285,13 +294,15 @@ func logsHandler(logger *slog.Logger, provider Provider) http.HandlerFunc {
 		defer func() {
 			// Send a close frame on the way out so the appview proxy
 			// sees a clean shutdown. Mirrors upstream
-			// spindle.(*Spindle).Logs.
+			// spindle.(*Spindle).Logs. WriteControl honours the
+			// deadline argument directly, so a stuck peer can't hang
+			// us here.
 			_ = conn.WriteControl(
 				websocket.CloseMessage,
 				websocket.FormatCloseMessage(
 					websocket.CloseNormalClosure, "log stream complete",
 				),
-				time.Now().Add(time.Second),
+				time.Now().Add(wsWriteWait),
 			)
 			conn.Close()
 		}()
@@ -334,6 +345,20 @@ func logsHandler(logger *slog.Logger, provider Provider) http.HandlerFunc {
 					)
 					return
 				}
+				// Bound the write so a client that stopped reading
+				// but kept the TCP connection open can't hang us on a
+				// full kernel send buffer. WriteMessage doesn't take a
+				// deadline argument the way WriteControl does — we
+				// have to set it on the conn before each frame.
+				if err := conn.SetWriteDeadline(time.Now().Add(wsWriteWait)); err != nil {
+					logger.Debug("logs set write deadline failed",
+						"err", err,
+						"knot", knot,
+						"pipeline_rkey", pipelineRkey,
+						"workflow", workflow,
+					)
+					return
+				}
 				if err := conn.WriteMessage(websocket.TextMessage, frame); err != nil {
 					logger.Debug("logs frame write failed",
 						"err", err,
@@ -440,9 +465,13 @@ func eventsHandler(logger *slog.Logger, br *broker) http.HandlerFunc {
 					return
 				}
 			case <-ticker.C:
+				// WriteControl takes its own deadline argument, so
+				// the ping itself can't hang us — but we still want a
+				// generous-but-bounded ceiling to match the per-frame
+				// write timeout.
 				if err := conn.WriteControl(
 					websocket.PingMessage, nil,
-					time.Now().Add(time.Second),
+					time.Now().Add(wsWriteWait),
 				); err != nil {
 					logger.Debug("events ping failed", "err", err)
 					return
@@ -475,6 +504,13 @@ func streamEvents(ctx context.Context, conn *websocket.Conn, st *store, cursor *
 		if err != nil {
 			return fmt.Errorf("marshal envelope: %w", err)
 		}
+		// Bound the per-frame write so a client that stopped reading
+		// (but didn't close the TCP connection) can't hang the
+		// handler on a full kernel send buffer. WriteMessage has no
+		// deadline argument of its own — we set it on the conn.
+		if err := conn.SetWriteDeadline(time.Now().Add(wsWriteWait)); err != nil {
+			return fmt.Errorf("set write deadline: %w", err)
+		}
 		if err := conn.WriteMessage(websocket.TextMessage, frame); err != nil {
 			return fmt.Errorf("write frame: %w", err)
 		}