Instance Incidents
Translation pending — the Turkish page is the source of truth.
Starting with v0.0.92, every error-boundary or pipeline-level failure on a workflow instance is recorded as a structured incident in its own InstanceIncidents table, with unlimited history and a denormalized HasActiveIncident flag on the instance. The state function, GET .../instances/{instance}, and the instance list all expose the same incident link block (hasActiveIncident, active.href, history.href) instead of embedding incident content, so clients follow a link to read the newest unresolved incident (GET .../incidents/active, 404 is a normal "nothing open" answer) or the full paged history (GET .../incidents). One failure now produces exactly one incident row, a retry that faults again keeps the instance durably Faulted, and a successful retry closes every open incident and recomputes the flag. See the Turkish page for the full field reference, retry semantics, and script-side context.Incident access.