There is an issue where a replica can connect to the primary's stream but not be notified when the stream shuts down. This means the replica will continue to wait until the connection times out. Some possible fixes include reducing the connection timeout, adding "ping" frames, and adding a "end" frame.
I added an END frame via https://github.com/superfly/litefs/pull/183 and it seems to be handling deploys and restarts better. I'm going to close this for now and can reopen if there's additional issues.
There is an issue where a replica can connect to the primary's stream but not be notified when the stream shuts down. This means the replica will continue to wait until the connection times out. Some possible fixes include reducing the connection timeout, adding "ping" frames, and adding a "end" frame.