`meetings/2026-05-07_090002-morning.md`

52 lines
# DW Morning Stand-Up — Thu 2026-05-07 09:00 PDT

---

## 1. Carryover from Last Meeting

1. ⏳ **george-gmail OAuth rotation** — still unrotated. Restart count: 12. Quiet stretch was process luck, not a fix. Next crash resets the clock.
2. ⏳ **Doctor apply-templates.js go/no-go** — 5+ days blocked on Steve's yes/no. Lawyer Iter 46 is behind it.
3. ⏳ **Gemini key #1 + #2 + PG dw_admin rotation** — runbooks ready, 32 min of Steve's time needed. Not touched.
4. ⏳ **Wallpaper cluster crashes** — overnight data shows the problem is much larger than 8 sites. See §2.

---

## 2. Overnight (Last 12 Hours)

**Crashes / instability** — fleet is not stable.

- `ventura-claw` (30 restarts), `ventura-claw-leads` (44), `ventura-corridor` (87) — these are the worst on the board; no fix has landed.
- `agedwallpaper`, `pastelwallpaper`, `mylarwallpaper`, `architecturalwallcoverings`, `greenwallcoverings`, `glitterwalls` — all at 21–22 restarts. The wallpaper cluster is 30+ sites crashing, not 8. This is a fleet-wide pattern, not site-specific bugs.
- `dw-collections-viewer` (29), `dw-universe` (15), `fleet-tv` (14), `claude-control-center` (13) — non-wallpaper crashers added to the pile.
- `national-paper-hangers` (11) — fix landed last night (route/middleware null read); appears stabilized but restart count is still elevated.

**Long-running jobs** — `graphic-loop` passes 5–8 ran overnight across the decade wallpaper sites (1930s–1980s). All show commits 15 min–11 hours ago; Josefin Sans / Playfair Display / Bungee Shade theme-fit corrections are landing. Pipeline is active and progressing.

**External signals** — no remote collaborator commits. NationalPaperHangers got e2e regression test additions 15 min ago (sitemap + robots.txt checks). All other git activity is local agent work.

---

## 3. Today's Plan + Agent Assignments

1. `pm2-crash-watcher` — pull logs for ventura-claw + ventura-corridor to identify root cause — restarts at 87/44/30 suggest a shared dependency or config issue, not random failures — error pattern identified (specific exception or missing env var)
2. `debugger-agent` — check if wallpaper cluster (30+ sites) shares a single server.js / PM2 config entry point — a single broken dependency would explain fleet-wide simultaneous restart counts — confirm shared vs. independent process trees
3. `Steve` — yes/no on Doctor apply-templates.js — 5 days blocked; lawyer directory Iter 46 is also gated — one decision unblocks two pipelines
4. `Steve` — run george-gmail OAuth rotation (10 min) — 12 restarts means the token has already expired and re-requested at least a dozen times — token rotated and restart count resets
5. `Freddy-port agent` — proceed with Freddy port rotation (soak window expired by tonight) — `_shared/TODO.md` says OK after 2026-05-07 evening — `rotate-pg-scoped.sh` completes without error
6. `graphic-loop` — continue passes on remaining decade sites — pipeline is clearly working; let it run — all decade sites show pass 8+ commits

---

## 4. Blockers & Decisions Steve Owes

- **Doctor apply-templates.js** — yes/no. Every day of delay is a day lawyer Iter 46 doesn't run.
- **george-gmail OAuth** — token rotation requires Steve's hands; can't be delegated. 10 minutes.
- **Gemini key #1 + #2 rotation** — paste via `/secrets`; runbook is ready. ~32 min total with PG dw_admin.
- **ventura-claw / ventura-corridor root cause** — decision needed: kill and rebuild vs. debug in place. 87 restarts suggests rebuild may be faster than log archaeology.

---

## 5. One Thing to Think About

The wallpaper cluster has grown from 8 sites to 30+, all crashing at similar restart counts. This looks less like per-site bugs and more like a shared infrastructure failure — same Node version, same PM2 config block, same missing env var, or a shared dependency that broke. Before spending time on individual site fixes, is there a single shared `server.js` template these all fork from? If yes, one fix propagates to all 30.