fix issue where channel was closing

src/replication/events.rs

@@ -1,6 +1,6 @@
 //! events related to replication
 use crate::{common::BitfieldUpdate, HypercoreError};
-use async_broadcast::{broadcast, Receiver, Sender};
+use async_broadcast::{broadcast, InactiveReceiver, Receiver, Sender};
 
 static MAX_EVENT_QUEUE_CAPACITY: usize = 32;
 
@@ -75,13 +75,19 @@ impl_from_for_enum_variant!(Event, Have);
 pub(crate) struct Events {
     /// Channel for core events
     pub(crate) channel: Sender<Event>,
+    /// Kept around so `Events::channel` stays open.
+    _receiver: InactiveReceiver<Event>,
 }
 
 impl Events {
     pub(crate) fn new() -> Self {
-        let mut channel = broadcast(MAX_EVENT_QUEUE_CAPACITY).0;
+        let (mut channel, receiver) = broadcast(MAX_EVENT_QUEUE_CAPACITY);
         channel.set_await_active(false);
-        Self { channel }
+        let mut _receiver = receiver.deactivate();
+        // Message sending is best effort. Is msg queue fills up, remove old messages to make place
+        // for new ones.
+        _receiver.set_overflow(true);
+        Self { channel, _receiver }
     }
 
     /// The internal channel errors on send when no replicators are subscribed,