Automatically continue a running session on a persistent environment when the client's environment goes offline #15609

What happened

This is a feature request, not a bug.

I use two T3 environments: my laptop (local) and a home lab (always on). Both have the same provider, model and authentication set up. When I start a session on the laptop and then close the lid, shut it down, or lose network, the session stops, because it is tied to the machine it started on.

I'd like T3 to move the session's context to the home lab automatically when the laptop goes offline, so the work keeps running without me intervening.

Diagnosis

Checked against the v0.0.45 source (6c8fed35). I found no cross-environment thread transfer or failover code. The existing "handoff" code only switches provider or model within one thread on one server (for example #12352 and #13882).

The environment is chosen when a thread is created. After the first message, the thread stays on that environment.

Manual transfer was proposed and attempted but never shipped:

• [Feature]: Transfer a provider-native session between T3 Connect environments #5363 was locked and converted to a discussion on 2026-08-15. It was not completed.
• PR feat(orchestrator): move a running thread between devices #5544 ("move a running thread between devices") was closed on 2026-09-08 as an abandoned draft. It was stacked on a branch with no common ancestor with main. It was not merged.

This request goes further than #5363. Automatic failover has to work after the source has gone away, so a manual "pause, copy, resume" is not enough. It would need:

• Session state available to the destination before the source dies (replication, or storage both environments can reach).
• Detection that the source environment is gone.
• A single-owner guarantee, so the laptop doesn't resume the same session when it comes back.
• A preflight check that the destination has the same provider, CLI version, authentication and project checkout.

Steps to reproduce

1. Configure two environments, a laptop and an always-on home lab, with the same provider and authentication.
2. Start a session on the laptop and send a message that starts a long-running turn.
3. Turn the laptop off or disconnect it.
4. Expected: the session continues on the home lab with the same history and provider-native continuation.
5. Actual: the session stops. There is no way to move it to the home lab, since the thread's environment is locked after the first message.

Version

0.0.45

Environment

Linux x64 (7.0.0-111031-tuxedo), Node v26.8.2, launched via npx t3 triage. Local server running on 127.0.0.1:3773.

Evidence

None. This is a missing capability, not a failure. A search of v0.0.45 for cross-environment thread transfer or failover code found nothing.

Related issues

• [Feature]: Transfer a provider-native session between T3 Connect environments #5363: manual session transfer between T3 Connect environments. Locked and converted to a discussion without a fix.
• feat(orchestrator): move a running thread between devices #5544: PR implementing manual transfer. Closed unmerged.

This request is not a duplicate. It asks for automatic failover to a persistent environment after the source goes offline, which neither of those covered.

Fix applied or workaround

None. The only workaround is to start a new thread on the home lab and summarize the context by hand.

Filed by

claude (claude-sonnet-5-5) via t3 triage