Two things need you, both about 2 minutes each. Everything else is done or teed up.
wrangler secret put commands. Until they're set, no alert can send.Also: make-give-live is still paused — you only restored Daily Quests.
Crazy Domains → tomwalkerelectrical.co.nz → Name Servers → delete both ns1/ns2.crazydomains.com → add these two → Save. DNSSEC off.
garret.ns.cloudflare.com hera.ns.cloudflare.com
His login is in the Gmail thread. .nz propagates in ~15 min–2 hrs, then it's live automatically — Cloudflare side is fully wired. Tell him after; he asked to know.
cd C:\Users\PC\Documents\supabase-watchdog npx wrangler secret put RESEND_API_KEY npx wrangler secret put DQ_SERVICE_KEY
Resend key from the Resend dashboard. DQ_SERVICE_KEY = Supabase → Daily Quests → Settings → API → service_role. MGL's key is already set.
The watchdog measured HTTP 530 on it at 00:29 UTC. Restore it from the dashboard (plain Restore, not "Restore to a new project" — that changes the ref and breaks the app).
Also worth exporting both DBs while they're up — the dashboard shows LAST BACKUP: No backups, and free tier has none.
I checked the Companies Register before applying and found a blocker. The active entity is:
| Legal name | JB SOLAR AND ELECTRICAL 2023 LIMITED |
| Company / NZBN | 8676206 · 9429051214975 |
| Also on register | JB SOLAR AND ELECTRICAL LIMITED (2017) — Removed |
Everything on file says "JB Solar & Electrical Ltd". Apple and Google reject org enrolments on exactly that — & vs and, Ltd vs Limited.
Confirm the illion D-U-N-S record (758610119) reads the exact registered name before applying, and allow ~2 business days for a fix to reach Apple/Google. Applying first means a bounce and weeks lost.
The story on the board was wrong. The keep-alive never stopped firing — Cloudflare analytics show 27 of 29 days, 1 run/day, 0 errors. Supabase paused it anyway on 25 Jul.
So the pings were never counted. An anonymous read that RLS filters to [] registers as nothing — your dashboard shows what Supabase actually measures: EGRESS 0GB · DB 0GB · MAU 0.
No keep-alive can ever work. That ends three rounds of redeploying the same thing.
Why it stayed invisible 8 days: a paused project keeps answering HTTP (530), so fetch() never throws and Cloudflare logs "success" — and the old worker only console.log'd its own failure. DNS teardown trails the pause by about a week, which is why my first read of it was backwards.
Doesn't try to prevent pausing — that's impossible. It warns you before it happens:
supabase-watchdog.josh-3f5.workers.dev · 6-hourly · repo jbsolar405/supabase-watchdog. First run correctly caught daily-quests UP, make-give-live PAUSED.
Zone created, both Pages domains bound, proxied CNAMEs live, site verified serving 200 with his final edits (Call Tom button, 173 Tara Road). Approved build is tideline-v2, not the older tideline-6jz.
You'd already replied 2 Aug, so the board's "no acknowledgement sent" was stale — corrected. Calendar reminder set Mon 11 Aug 8am to chase Colm for the 5kW Solis ETA and book her in.
Her dates: any Thursday in August, or Mon 17 Aug.
You were right that one account holds multiple businesses. What I'd flagged was narrower: renaming the verified JBSE listing to Evoca would convert it and cost you the reviews and ranking history.
Plan: leave JBSE alone, add a second profile for Evoca. It'll need its own video verification (unedited, 30s+, showing location, branding and a document proving authority). Tidy the Evoca branding first — same address as an existing listing gets extra scrutiny.
The 12-testers / 14-day closed-test requirement applies only to personal accounts. Organization accounts publish straight to production — that's ~3 weeks off every Android launch, and a solid reason to go org.
Heads up: Apple publishes the legal entity as the App Store seller, so apps show "JB Solar and Electrical 2023 Limited", not Evoca. Play's public developer name is free text, so Evoca can show there.
Left the old daily-quests-keepalive worker in place rather than deleting it — deletes are ask-first. It's harmless but should go, since a dead cron implies protection it never gave.
Board + memory both corrected and pushed: the "3rd silent keep-alive failure" story, and my own earlier wrong claim that a pause removes DNS instantly.
Left alone as you said: Make Give Live (handover), Gavin. JA Russell's GoodWe quote arrived 28 Jul — that was Delwyn's alternative and she picked the Solis, so nothing owed.