diff --git a/CHANGELOG.md b/CHANGELOG.md
index 34198fa0f..6593bae75 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -44,7 +44,7 @@
 - Skills: add tmux skill + interactive coding guidance in coding-agent.
 - Gateway: document port configuration + multi-instance isolation.
 - Onboarding/Config: add protocol notes for wizard + schema RPC.
-- Queue: clarify steer-backlog behavior and update examples for streaming surfaces.
+- Queue: clarify steer-backlog behavior with inline commands and update examples for streaming surfaces.
 
 ## 2.0.0-beta5 — 2026-01-03
 
diff --git a/docs/queue.md b/docs/queue.md
index 518db7de4..aafda0ceb 100644
--- a/docs/queue.md
+++ b/docs/queue.md
@@ -30,6 +30,7 @@ Inbound messages can steer the current run, wait for a followup turn, or do both
 Steer-backlog means you can get a followup response after the steered run, so
 streaming surfaces can look like duplicates. Prefer `collect`/`steer` if you want
 one response per inbound message.
+Inline fix: `/queue collect` (per-session) or set `routing.queue.bySurface.discord: "collect"`.
 
 Defaults (when unset in config):
 - All surfaces → `collect`
diff --git a/src/agents/pi-embedded-subscribe.ts b/src/agents/pi-embedded-subscribe.ts
index e1b76efca..268d5c9d1 100644
--- a/src/agents/pi-embedded-subscribe.ts
+++ b/src/agents/pi-embedded-subscribe.ts
@@ -259,6 +259,7 @@ export function subscribeEmbeddedPiSession(params: {
     if (!blockChunking) return;
     const minChars = Math.max(1, Math.floor(blockChunking.minChars));
     const maxChars = Math.max(minChars, Math.floor(blockChunking.maxChars));
+    // Force flush small remainders as a single chunk to avoid re-splitting.
     if (force && blockBuffer.length > 0 && blockBuffer.length <= maxChars) {
       emitBlockChunk(blockBuffer);
       blockBuffer = "";
@@ -457,6 +458,7 @@ export function subscribeEmbeddedPiSession(params: {
               if (delta) {
                 chunk = delta;
               } else if (content) {
+                // Providers may resend full content on text_end; append only the suffix.
                 if (content.startsWith(deltaBuffer)) {
                   chunk = content.slice(deltaBuffer.length);
                 } else if (deltaBuffer.startsWith(content)) {