Distributed Coordination
Forum does not provide a replicated coordination or claim protocol.
The current Forum model has no transport client, remote-log observer, index replay/import method, lease service or consensus protocol. Replicating its underlying Strand bytes does not update another Forum instance’s live_tuples map. The SDK wrappers open local named stores; they do not add that missing coordination layer.
A take result is scoped to the map in that Forum instance. It does not authorize a cross-process claim. A Lens read-then-write flag is also not an atomic distributed claim and cannot establish exactly-once work execution.
For an application that needs durable work claims, first implement or connect an authoritative claim service with atomic state transitions, recovery and idempotent external effects. Keep Forum operations serialized locally while the core lock-order and recovery gaps remain. Do not infer eventual tuple convergence or a latency bound from the presence of Strand replication APIs.
