Workflow: continue a durable Hub execution without a second control channel

[timigod]

What I am trying to do

An external controller dispatches durable work to enrolled, independent Paseo daemons. Hub already gives each create a durable executionId, so the controller can safely retry after a lost response or daemon restart without creating a duplicate agent.

After that agent becomes idle, the controller sometimes needs to continue the same owned execution with another prompt.

Current workflow

1. Create through Hub with a controller-owned executionId.
2. Receive the agent mapping and updates from Hub.
3. To continue, open a separate authenticated DaemonClient connection, resolve the agent again, and send the prompt there.
4. Keep execution-finality and release policy in the external controller.

The separate channel works, but it splits one durable execution across two control identities. Hub create is retry-safe; the continuation path must independently reconstruct ownership and deduplication.

Where the workflow gets in the way

A controller can safely replay Hub creation, but it cannot safely express “send this next operation to the execution I already own” through the same durable identity. Generic message sends can carry a client message ID, but Hub does not currently expose a continuation operation keyed by executionId.

Desired flow

Would Hub be open to one generic continuation operation that:

• addresses the existing execution by executionId;
• carries a caller-supplied operation/message deduplication key;
• returns the current mapped agent snapshot or a precise ownership/lifecycle error;
• preserves Hub’s daemon-scoped authority; and
• leaves execution-finality policy outside Paseo?

This proposal does not ask Paseo to own multi-host discovery, fleet aggregation, release receipts, deployment topology, or a global controller registry. It is only about continuing one already-owned durable Hub execution without switching to a second control channel.