Specific data streams can cause Bitbucket nodes to become unresponsive.
The following can be found in the logs:
WARN [hz.hazelcast.IO.thread-Acceptor] c.h.nio.tcp.SocketAcceptorThread [<ip>]:5701 [<cluster name>] [3.7.4-atlassian-43] java.io.UTFDataFormatException: Rejecting request to read 16908289 UTF characters
Leading to this:
2019-02-21 02:22:53,698 WARN [hz.hazelcast.cached.thread-5] com.hazelcast.nio.tcp.TcpIpAcceptor [<ip>]:5701 [<cluster name>] [3.11.1] com.atlassian.stash.internal.cluster.NodeConnectionException: Cannot verify whether the nodes connect to the same database and shared home
Block any traffic that is not from another Bitbucket node on port 5701 (i.e. set up a firewall).