Vert.x 4 Cluster Manager transition to modular discovery mechanisms

Vert.x 4 has shifted toward a more modular approach for selecting the underlying discovery mechanism within the Cluster Manager, moving away from the tightly coupled defaults of previous versions. When scaling the Event Bus across multiple JVMs, the choice of cluster manager directly impacts how nodes discover each other and maintain shared state.

There is uncertainty regarding the behavioral differences when transitioning from the standard Hazelcast-based discovery to alternative modular managers in high-latency network environments. Specifically, the impact on message routing consistency and the risk of split-brain scenarios varies depending on the discovery implementation used.

• Does the modular discovery configuration in Vert.x 4 alter the default timeout behaviors for node heartbeat detection compared to the legacy Hazelcast integration?
• What are the specific configuration constraints when implementing a custom cluster manager to ensure the Event Bus maintains strict request-response delivery guarantees?