Akka Cluster Serialization Binding Incompatibility with Akka Persistence Journal During Rolling Upgrade

The goal is to determine whether persistent actors can recover journal state when serializer bindings differ between upgraded and legacy nodes in an Akka Cluster undergoing a rolling upgrade. Incompatible serializer versions may trigger member exits before recovery completes, preventing persistent actors from accessing journal state.

Journal replay may stall if the underlying journal plugin schema changes between Akka versions, as no automatic backward-migration exists for snapshot or event formats. CoordinatedShutdown interactions with persistent timers and snapshot cleanup are version-sensitive, potentially leaving actors in recovery limbo when nodes operate beyond the supported rolling upgrade window.

What conditions cause serializer version mismatches to trigger member exits before persistent actor recovery completes? Which serialization binding configurations prevent silent data loss when rolling Akka cluster versions across persistent actors? Can cluster sharding entity passivation be automated when journal schema changes between Akka versions without manual state re-initialization?