diff --git a/src/structures/session.ts b/src/structures/session.ts
index 505df7e75..86e81d68b 100644
--- a/src/structures/session.ts
+++ b/src/structures/session.ts
@@ -853,6 +853,8 @@ export default abstract class Session {
                 round.songStartedAt = Date.now();
             }
 
+            await this.ensureConnectionReady();
+
             this.connection.play(stream, {
                 inputArgs,
                 encoderArgs: Object.entries(encoderArgs).flatMap((x) => [
@@ -901,13 +903,13 @@ export default abstract class Session {
                 }
             }
 
+            this.stopGuessTimeout();
+
             if (clipAction === ClipAction.END_ROUND) {
                 // The end round clip doesn't deal with round state, it just plays and ends
                 return;
             }
 
-            this.stopGuessTimeout();
-
             if (this.isGameSession() && this.isClipMode()) {
                 const clipGameRound = round as ClipGameRound;
                 if (!round.finished) {
@@ -1319,4 +1321,28 @@ export default abstract class Session {
             );
         });
     }
+
+    private async ensureConnectionReady(): Promise<void> {
+        if (!this.connection) {
+            throw new Error(
+                "Connection is unexpectedly null in ensureConnectionReady",
+            );
+        }
+
+        if (!this.connection.piper.encoding) {
+            return;
+        }
+
+        logger.warn("Connection is unexpectedly in encoding state.");
+        await delay(1000);
+
+        // if still encoding, force stop
+        // eslint-disable-next-line @typescript-eslint/no-unnecessary-condition
+        if (this.connection.piper.encoding) {
+            logger.warn(
+                "Connection is still in encoding state after timeout, force stop.",
+            );
+            this.connection.stopPlaying();
+        }
+    }
 }