Backend fix is LIVE — the alerts are still firing because the monitors haven't been repointed yet. Code deployed a dedicated pure-health endpoint (static 200, no DB — can't time out) + a 5-min warmup cron. Action needed (UptimeRobot dashboard, ~2 min): (1) point each flapping monitor at https://mbnakfrkfwxinhxlisyq.supabase.co/functions/v1/pure-health (keyword monitor on "ok":true); (2) raise timeout to 30s; (3) require 2 consecutive fails before alerting. Until the monitors are repointed they still hit the old business RPCs, so the texts continue — this is the last, manual step.