There might be a long laundry list of things one might want to consider prior to deploying to production. Remember, now that we are using persisted events as the fundamental elements to hydrating our domain, we need to pay special attention to how we persist them.
Event Journal Database
Since events stored per aggregate are read in a paginated way, it makes sense to chose a persistence that supports this kind of query well. SQL data bases are fine also, just make sure you do not ship with the defaul akka inmemory persistence plugin. As you can see there are many plugins (1,2,3,4). Doing some research here is crucial to match your needs.
Test all your models to make sure that they are serializable and deserializable. This will give you good assurances on production usage, especially in clustered scenarios. The reason for this is that within the same local actor system, messages are passed as normal CLR objects. When you have clustered systems doing communications, then these messages become serialized. If you are planning to run your actor system in a clustered scenario you should invest in testing all the messages that may go across the wire.
Clustering Seed Node
This is applicable to those who plan to deploy in a clustered environment. It would be most advantageous if you had a seed node in your cluster to facilitate the cluster gossip. 1 or more well known addresses that can be used. In windows, Lighthouse attempts to help you out here by hosting the seed as a windows service and exposing the required ports. Our clustered example seed here, is actually good enough for many scenarios, the seed node need not even play a part in your applications deployment considerations. The seed is a statically located akka actor system, part of a cluster, that can speak the same gossip language as your worker/client actor system.