Operator gate — flips status pending → active. Until activated, the intake worker skips the source.

Also resets consecutive_failures to 0 and clears last_error on transition. Without this reset (pre 2026-05-31), re-activating a previously-auto-disabled source would carry the old failure streak over — one more failure would immediately re-auto-disable it. Worse, if status was manually flipped active by a sibling code path that DIDN'T touch cf, the row sat at status=active, cf>threshold forever and burned poll cycles without ever firing the disable gate (since the threshold check runs only at the cf++ moment). See triage /tmp/storyflo_intake_log_triage_2026_05_31.md.