Choose or clear Slack's target dataset. Owner-only (CLO-1014).

payload.dataset_id is checked against :func:~.flow.list_writable_datasets, never trusted from the request body alone — that list is also where a name shaped like some connector's own default has already been excluded (finding #1), and this route stores that list's own name for the id, never payload.dataset_name (finding #2), same as the generic connector framework's version of this route.

delete_old_target is far more limited here than in the generic framework. Slack's bulk messages are tagged (modules/integrations/slack_dlt_source.py's SOURCE_TAG), but the attachments a sync also ingests go through a plain, untagged remember() call (get_slack_sync_router.py's _ingest_attachments) — there is no way to identify which rows in a shared dataset are this connection's attachments, so leaving a chosen target always refuses the delete (409, finding #11) rather than leaving them orphaned or, worse, guessing. Leaving Slack's own default is different: nothing else can ever share that name (f"slack-{team_id}" is exclusive to this connection by construction, the same invariant dataset_name_for relies on for every other connector's default), so an unscoped whole-dataset wipe is safe there and is allowed.

Write, then delete, never the other way round (finding #6) — same ordering and the same reasoning as the generic framework's version: the new target is recorded first, under the row's own lock with an optimistic check against a concurrent write, and only once that has committed does a requested delete run.

sweep_previous_only (CLO-1069) switches this into a dedicated mode that never resolves or writes a new target at all - see the generic framework's own field docstring and its _sweep_previous_ only for the retry-banner bug this closes.