fix(rebuild): the rebuild script deletes a restarted hub's fresh handshake, then exits 1 after a successful deploy #1797

Important

Problem — On 2026-10-13, rebuilding the Mac Studio (a hub machine) from main with the terminal rebuild script to deploy #1781 hit two defects. Both happened after the deploy itself had succeeded.

Defect 1: the script deletes the restarted hub's handshake. When the engine overlay has changed, the script stops the running engine and then deletes the handshake record unconditionally. On a machine whose hub runs as a launchd KeepAlive service, launchd has already restarted the hub, and within about a second the new hub writes a fresh handshake. The script then deletes that new one. On the Studio the hub came up at 17:15:47, wrote its handshake, and the record was gone at 17:15:48. The hub writes its handshake only at startup, so every window on the machine is left without one. A window reload then finds the engine's port in use with no handshake, takes the reclaim path, and restarts the hub at an uncontrolled moment. Recovering took a second, manual hub restart.

Defect 2: the script exits 1 after a successful deploy. The step that rehomes sessions from deleted worktrees ends on a conditional that is false when nothing needed rehoming, so the function returns non-zero. Under the script's exit-on-error setting the script stops there, after the deploy has completed but before its final summary, and exits 1.

Approach — Delete the handshake only while it still names the process the script stopped. Give the rehome step an explicit successful return.