We'll need to be careful to ensure that any "downstream" messages (ie node_announcement/channel_update) to a pending channel_announcement are held and not processed until the async query resolves. Ideally we can do this without storing the pending messages in the network graph itself, maybe just storing a map from the pending SCIDs to a weak pointer to the future, adding pending to-process messages to the future object itself and making users call back into the graph with the future object.
We'll need to be careful to ensure that any "downstream" messages (ie node_announcement/channel_update) to a pending channel_announcement are held and not processed until the async query resolves. Ideally we can do this without storing the pending messages in the network graph itself, maybe just storing a map from the pending SCIDs to a weak pointer to the future, adding pending to-process messages to the future object itself and making users call back into the graph with the future object.