Slack: a stream segment that Slack expired is left showing "Something went wrong" #1036

Bug Description

When Slack expires a native stream segment before the adapter stops it, stream() either continues in a new message or returns the expired one, and never edits the expired message. Slack shows a stream it closed itself as a "Something went wrong" pill. The thread keeps an error-looking message even though the reply was delivered.

Code on main (1e7a61e), packages/adapter-slack/src/index.ts:

• The rotateSegment catch path ("stream segment expired before rotation, continuing in a new message") starts the next segment and leaves the expired one as is.
• The final stop() catch path ("stream-end blocks skipped" / "delivering the rest in a new message") returns or abandons the expired message without editing it.
• The only editMessage in stream() belongs to the post-and-edit fallback.

Steps to Reproduce

1. createSlackAdapter({ agentView: true }) with the default streamSegmentMaxAgeMs.
2. thread.post(source), where source yields some text, stays silent for 6 minutes (a long tool call), then yields more.
3. Slack expires the first segment during the gap. On the next delta, rotateSegment's stop() fails with message_not_in_streaming_state, and the reply continues in a new message.
4. Look at the first message.

Expected Behavior

The expired message keeps the content it already streamed, like a normally stopped segment.

Actual Behavior

The expired message stays as Slack sealed it, and Slack renders that as "Something went wrong". We found this by reading the code. The rendering comes from our direct API test and the reports below, not from a run through the adapter.

Chat SDK Version

4.41.0 (also on main @ 1e7a61e)

Platform Adapter

Slack

Additional Context

Evidence that a sealed, never-stopped stream renders as an error, and that chat.update repairs it:

• Our direct API test (28 Sep 2026, channel thread, plan mode): an append at ~3m45s worked. At ~9m25s, chat.appendStream and chat.stopStream both returned message_not_in_streaming_state, and the message showed "Something went wrong" in place of the plan.
• message_not_in_streaming_state after undocumented timeouts hit slackapi/python-slack-sdk#1859 reports the same pill. Slack's fix, feat(web): expose public ts property on ChatStream slackapi/python-slack-sdk#1880, exposes the stream ts so apps can call chat.update, which it calls the only recovery path.
• slack: native streams expire after about five minutes geminixiang/mikan#157: chat.update on an expired stream message succeeds, and users see the final answer.
• fix(slack): keep progress updating after native streams expire openclaw/openclaw#141170: a real-API readback shows chat.update rewriting the same message, tasks included, after message_not_in_streaming_state.

Suggested fix: in both expired paths, chat.update the expired message (segment.streamer.ts) with the content it already had, close any open task, then continue as today.

Related: #911 (a rotated segment with an in-progress task shows a red error icon; still on main, since rotateSegment's stop() sends no closing task_update), #884, #607.