diff --git a/roomserver/internal/input/input.go b/roomserver/internal/input/input.go
index acffde5d..1eab6780 100644
--- a/roomserver/internal/input/input.go
+++ b/roomserver/internal/input/input.go
@@ -54,8 +54,10 @@ type Inputer struct {
 func (r *Inputer) Start() error {
 	_, err := r.JetStream.Subscribe(
 		r.InputRoomEventTopic,
+		// We specifically don't use jetstream.WithJetStreamMessage here because we
+		// queue the task off to a room-specific queue and the ACK needs to be sent
+		// later, possibly with an error response to the inputter if synchronous.
 		func(msg *nats.Msg) {
-			_ = msg.InProgress()
 			roomID := msg.Header.Get("room_id")
 			defer roomserverInputBackpressure.With(prometheus.Labels{"room_id": roomID}).Dec()
 			var inputRoomEvent api.InputRoomEvent
@@ -65,17 +67,22 @@ func (r *Inputer) Start() error {
 			}
 			inbox, _ := r.workers.LoadOrStore(roomID, &phony.Inbox{})
 			inbox.(*phony.Inbox).Act(nil, func() {
-				_ = msg.InProgress()
 				if err := r.processRoomEvent(context.TODO(), &inputRoomEvent); err != nil {
 					sentry.CaptureException(err)
-					_ = msg.Respond([]byte(err.Error()))
 				} else {
 					hooks.Run(hooks.KindNewEventPersisted, inputRoomEvent.Event)
 				}
 				_ = msg.Ack()
 			})
 		},
+		// NATS wants to acknowledge automatically by default when the message is
+		// read from the stream, but we want to override that behaviour by making
+		// sure that we only acknowledge when we're happy we've done everything we
+		// can. This ensures we retry things when it makes sense to do so.
 		nats.ManualAck(),
+		// NATS will try to redeliver things to us automatically if we don't ack
+		// or nak them within a certain amount of time. This stops that from
+		// happening, so we don't end up doing a lot of unnecessary duplicate work.
 		nats.MaxDeliver(0),
 	)
 	return err