basquetwi
BWI-51
Auto-close open questions in the transaction that closes an item
Done normal
unassigned
Makes the doctor check open_questions_on_closed_items unreachable rather than swept: closing an item would close its open questions in the same transaction and write a questionClosed/questionAnswered event carrying the reason. Belongs in the service like every other invariant - a client cannot skip it there. Open design question: silent auto-close loses the distinction between 'answered' and 'overtaken by events', so the event body must record which. Proposal awaiting Elazar's go; paired with the orphan-question backfill, which is the existing 6.
Activity
-
awaiting Elazar decision; raised on CLI 2026-08-16, surfaced by monitor-context-cc
-
3b shipped 2026-08-19: closing an item now closes its open questions in the same transaction, with the reason REQUIRED - the 400 names the open questions so the refusal is actionable. Two defects found while building it and fixed in the same commit: closedAt was re-stamped on every write to an already-closed row, and the CLI auto-filled the status body with the placeholder 'status=done', which silently satisfied the new guard.
-
terminal event backfilled: this item was closed via wi unblock before that route emitted one (defect found by wi doctor closed_items_missing_terminal_event, fixed same day)
task
2026-08-19 by wi-cli-venus
2026-08-19
2026-08-19 01:16