[sentry-junior]

Follows up on @billyvg's review of #127881.

Before this change, edit_thread_message called chat.update without a lifecycle or error handling. So when one answer's edit failed (for example, message_not_found after someone deleted the message), the error left the loop before the later answers were edited. That answer's pr_status_hash was never updated, so every later refresh failed on the same edit, and the latest answer's footer could stay on "Creating a PR in …".

What changed:

• edit_thread_message now runs inside its own SlackEntrypointEventLifecycleMetric (edit_thread_message).
• A missing message or channel (message_not_found, channel_not_found, is_archived) ends the edit. SlackIntegration.update_message raises IntegrationResourceNotFoundError for these, edit_thread_message records a halt and returns, and on_agent_update caches the new hash so later refreshes don't call chat.update on that reply again.
• Every other Slack error still raises, so refresh_agent_run retries the edit. Halt categories that can clear, like ratelimited or internal_error, are recorded as halts and raised. Other errors are recorded as failures and raised. A timeout is raised as IntegrationError.
• on_agent_update catches edit errors one message at a time. It still edits and caches the remaining answers, then raises the first error so the refresh retries only the edits that failed.
• Billy also asked when send_thread_update returns no message_ts. In practice that's a halt, which its lifecycle already records, but the record didn't say which run it was for. on_agent_update now logs seer.entrypoint.slack.agent_update.reply_not_posted with the run id and reply key. A halted post isn't retried: send_thread_update doesn't raise, so the reply waits for the run's next Seer completion hook. Retrying transient post halts would change send_thread_update, which process_thread_update and on_trigger_agent_error also use, so it's left out of this PR.

The shared SLACK_SDK_HALT_ERROR_CATEGORIES list is unchanged. message_not_found is now a recognized category, and only update_message treats it, along with the channel errors, as not found. Its other callers catch IntegrationError, which IntegrationResourceNotFoundError subclasses, so they still record a halt.

One test runs refreshes until nothing is left to edit. It covers a rate-limited edit that's retried without blocking the next answer, and a deleted message that halts once and isn't edited again. The reply_not_posted log has no test.

--

View Session [Sentry]

[joseph-sentry]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit ed6e16e. Configure here.}

[billyvg]

🤖 AI-assisted review, posted on behalf of @billyvg.

[billyvg]

🤖 AI-assisted review, posted on behalf of @billyvg.

[alexsohn1126]

looks good, thanks!