[Bug]: A parent is woken twice with the same delegated task result (after a restart cuts the delivery run, or when a follow-up repeats it) #32279

Before submitting

• I searched existing issues and did not find a duplicate.
• I included enough detail to reproduce or investigate the problem.

Area

apps/server

Steps to reproduce

Case A, restart:

1. A parent waits on delegate_task; the child completes; the delivery run starts on the parent and the agent reads the result, writes a reply and calls tools.
2. Restart the server before that run finishes.
3. After the restart, the same completion is delivered again.

Case B, repeated or unlabelled follow-up:

1. A child returns its result, then runs again on its own (a background-command wake) and replies with the same text, or with a new text.
2. The parent receives a second "Delegated task () [completed]" with no sign that it follows the first task.

Deterministic: in DelegatedCompletionDelivery.test.ts terms: (A) cut a delivery run that already wrote assistant output and finalize it; the task returns to pending and is re-sent. (B) start a child follow-up whose reply equals the last result; the parent gets a second delivery.

Expected behavior

A: a delivery the agent already answered counts as delivered, even if a restart cuts its run. A delivery run that started but wrote nothing is still re-offered, as #15323 intends.
B: a follow-up wake names the task it follows ("..., a follow-up of task after the child ran again"), and a follow-up the child started on its own whose reply repeats the last result is recorded (readable with task_status) without waking the parent. A follow-up started by a parent send or a user message always wakes.

Actual behavior

A: finalizeDelegatedCompletionDelivery returns every task of a cancelled delivery run to pending, so a restart that cut a run the agent had acted on sends the same completion again. B: the parent gets a second wake that looks identical to the first. Agents reacted with "This is the same review result again"; five parents saw a repeat between 24 Oct and 6 Nov.

Impact

Minor bug or occasional failure

Version or commit

0.1.8-nightly.20261106.4262 (c55db3e); code read at main 6577c6b, unchanged at 240446e.

Environment

macOS and Linux; Claude and Codex parents and children.

Logs or stack traces

None kept beyond the agents' replies; the deterministic cases fail on main.

Workaround

None; agents usually notice and ignore the repeat, at the cost of a turn.

Proposed fix (for the PR, once triaged)

A cut delivery run counts as delivered when it wrote answer items (assistant message, reasoning, tool calls). The follow-up wake label names the original task. A self-started follow-up whose reply equals the last result does not wake the parent. Follow-up command ids are deterministic so restart replays do not duplicate. Tests: "a delivery its run already answered is not sent again when the run is cut", "a restart that cuts an answered delivery does not send it again", "a follow-up wake names the task it follows", "a follow-up repeating the last result does not wake the parent" (fail on main); guards "a delivery cut before its run started is sent again once" and "a parent-sent repeat still wakes" pass on both.