Bug Description

All cron job deliveries fail in Telegram DMs with:

GrammyError: Call to 'sendMessage' failed! (400: Bad Request: message thread not found)

Environment

• OpenClaw: 2026.2.6-3
• Channel: Telegram (DM, not group/forum)
• Cron: isolated session + delivery.announce

Root Cause

Related to #10926. Commit 19b8416a8 introduced a regression where message_thread_id=1 is sent to Telegram API for DM chats. The fix in buildTelegramThreadParams() only skips thread_id=1 when scope === "forum", but DMs have scope === "dm".

What I tried (hotfixes, none worked)

1. Patched buildTelegramThreadParams() in all 3 dist files (extensionAPI.js, reply-B_4pVbIX.js, loader-n6BPnYom.js) to skip thread_id=1 unconditionally
2. Patched subagent-announce.ts to skip threadId="1" in callGateway params
3. Tried delivery.bestEffort: true
4. Tried sessionTarget: main + systemEvent

None resolved the issue because the thread_id appears to be injected at the gateway routing layer, after these functions.

Reproduction

{
  "schedule": {"kind": "at", "at": "<2min-from-now>"},
  "payload": {"kind": "agentTurn", "message": "test", "model": "haiku"},
  "sessionTarget": "isolated",
  "delivery": {"mode": "announce", "channel": "telegram", "to": "<dm-chat-id>"}
}

Expected

Message delivered to Telegram DM.

Actual

Infinite retries, each failing with thread_not_found. Sub-agent runs 20+ times burning tokens.

Impact

• All cron notifications to Telegram DMs are broken
• Each failed cron spawns 20+ retry sessions (token waste)
• Only workaround: sub-agent uses message tool directly with chatId (unreliable, causes spam)