memory: snapshot — Anaheim tunnel at 25% of circuit; selene retired; hrafn CI fixed

Session captured for a context reset. Six new detail entries.

THE OPEN ITEM: Anaheim's IPsec tunnel delivers ~550 Mbit/s aggregate against a
circuit measured at 2,153 Mbit/s. Not WireGuard (it is IPsec on ana-gw), not
CPU (idle), not crypto exhaustion (NPU-offloaded), not the fibre. Both tunnels
negotiate aes256-sha1; AES-GCM proposed. Operator signalled authorization;
execution pending, untracked by operator choice.

Also recorded: selene retired after losing a head-to-head on its own job with
chat-judge moved to gen and the model name left to 404; the 7-alias collision
on the gen seat that makes cross-alias corroboration an echo; hrafn adopted and
its CI found to have been reporting green while deploying nothing for its whole
life; all three Worldtree instances de-armed from a 69-day-stale :latest and
the Matrix homeserver re-plumbed to personal; every secret-bearing .env on
ana-docker tightened to 0600; the pfi org closing the repo-creation half of the
credential-migration directive; booth kept-board deletion and link pruning.

Two entries under Tried and abandoned: the CI checkout assertion that broke the
pipeline twice and was removed, and my proposal to alias a retired model name
at a different model, which the operator correctly overruled.

Index 271 -> 282 lines, under the 300 cap, so no archival fired.

persistent-memory.md

@@ -1,6 +1,6 @@
 # Persistent memory — eshpfi-management
 
-_Last updated: 2026-08-22_
+_Last updated: 2026-08-23_
 
 > **Always check for `/tmp/infra-ops-handoff.md`** — if it exists and its
 > `Written:` stamp is under an hour old, read it (it carries the in-flight
@@ -108,24 +108,32 @@ no longer deployed sidecars here. See Recent decisions.)
   (no NOPASSWD)** — stage model pulls to `/home`, not root-owned `/worktank`.
 ## Current state / in-flight
 
-_As of 2026-08-22 — three AI seats live on ana-ml2; `sec` is the one that moved this session._
+_As of 2026-08-23 — a long multi-party ops session; the one thing waiting on the operator is the Anaheim tunnel cipher._
 
-- **SEAT MAP.** **`gen`** = `orcarouter/Qwen3.8-27B-Uncensored` NVFP4-mixed, GPU0 :8015, 7 aliases, **still on the OLD nightly `311b3513` with MTP k=3**. **`char-rp`** = MeroMero-v2 dual-mode (prose + streaming CoT, one weight set, two aliases), GPU0 :8016, pinned `v0.26.0`. **`sec`/`sec-reasoning`** = M.O.G.-SEC, GPU1 :8019 — **rebuilt this session, see below**.
+- **⏳ THE OPEN ITEM: Anaheim's IPsec tunnel delivers ~25% of a verified 2 Gbps circuit.** Circuit measured at **2,153 Mbit/s**; the tunnel ceilings ~550 Mbit/s aggregate, ~230 single-stream. FortiGate CPU **idle**, IPsec **NPU-offloaded**, interface error-free — so it is not crypto exhaustion. Both tunnels negotiate **`aes256-sha1`**; **AES-GCM is the proposed change** and the operator has signalled he will authorize it. Not executed: production edge, needs a matching change at NH3 + ESH, each tunnel drops during renegotiation. → `persistent-memory.d/2026-08-23-anaheim-ipsec-tunnel-ceiling.md`
 
-- **🟢 `sec` NOW RUNS DFLASH2 ON A NEWER vLLM — promoted to its compose stack after real-use testing.** `nightly-e9d1398d` (+259 commits over production, `behind_by=0`), `dflash` k=7 with the 3.85 GB drafter, **util 0.52 / max-model-len 420,000 / KV ~453k**, 2048² vision. `restart: unless-stopped`, survives reboot. Canonical in `stacks/mog-sec/` with a fully-commented `.env.example`. **ROLLBACK:** `.env.bak-pre-dflash2-20260822` on the host, or swap `MOG_SPEC_CONFIG` + `MOG_IMAGE`. ⚠ `PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True` was **deliberately dropped** — the validated container never had it.
+- **🟢 SEAT MAP (unchanged this session except selene).** **`gen`** = `orcarouter/Qwen3.8-27B-Uncensored` NVFP4-mixed, GPU0 :8015, **now 7 aliases** (see the collision note). **`char-rp`** = MeroMero-v2 dual-mode, GPU0 :8016, pinned `v0.26.0`. **`sec`/`sec-reasoning`** = M.O.G.-SEC on DFlash2, GPU1 :8019. **`selene` RETIRED** — 17.2 GiB reclaimed on GPU1 (free now ~19.4 GiB).
 
-- **⚠️ THE `sec` DEGENERATION QUESTION IS OPEN AND CONFOUNDED.** It no longer degenerates, but **engine and drafter changed together**. **The isolating experiment is MTP k=3 on `e9d1398d`** — not yet run. Operator ruling: the degeneration lives in the **un-fixed vLLM**, not the weights; my MTP-head hypothesis is **retracted**. ⚠⚠ **Both the "degenerates at 2k" and "coherent to 10k" sightings are n=1 from uncontrolled sessions and are NOT evidence.** Production is **172 commits behind GDN spec-decode fix #53077**; `#51113` is present in both builds and is **necessary but insufficient**.
+- **⚠️ THE `sec` DEGENERATION QUESTION IS STILL OPEN AND CONFOUNDED.** Engine and drafter changed together; **the isolating experiment is MTP k=3 on `e9d1398d`** — still not run. Operator ruling stands: degeneration lives in the **un-fixed vLLM**, not the weights; the MTP-head hypothesis is **retracted**. Both prior sightings are n=1 and are NOT evidence. `gen` remains on the old nightly, untouched, gated on that experiment.
 
-- **⏳ `gen` IS UNTOUCHED and still on the old build.** If DFlash2 + the newer engine are the answer, gen is the obvious next beneficiary — but that decision is gated on the isolating experiment above, not on sec's n=1 result.
+- **🟢 ana-ml2 now mounts `/mnt/smithy`** (nh3-nas) **ro + soft, NOT in fstab** — needs a manual remount after reboot. For brokkr's R47 CPU work. Reads 24.7 MB/s sequential vs 98.3 on nh3-dev (that gap is the tunnel above), but **45 files/s vs 34** — small-file work is genuinely faster there. → `persistent-memory.d/2026-08-23-smithy-mount-ana-ml2.md`
 
-- **🟢 ESH IS DUAL-STACK; the v4 static is a Cityside ticket.** IPv6 live on `esh-userland` (SSID `PVC`) and `esh-server` from a delegated `2607:73c0:402:1d00::/56`; hosts egress over v6 as themselves, un-NATted. **v4 remains CGNAT (`100.104.3.250`) and a full gateway reboot proved the purchased static is NOT provisioned** — carrier ticket, nothing left to try locally. v6 firewall audited: default-deny inbound both versions, correct. NH3 stays v6-off deliberately (single /64 reserved for meshing). Flat-zone lateral-movement finding **parked, id 44**.
+- **🟢 ESH IS DUAL-STACK; the v4 static is a Cityside ticket.** IPv6 live on `esh-userland` and `esh-server` from a delegated `/56`. v4 remains CGNAT and a full gateway reboot proved the purchased static is **not provisioned** — carrier ticket, nothing left locally. NH3 stays v6-off deliberately. Flat-zone lateral-movement finding **parked, id 44**.
 
-- **🟢 OTHER SERVICES.** speaches ASR live irv-ml1:8204 (Eyra; loop closed). Open WebUI on esh-docker-vm:3211 (admin creds + admin-scoped API key vaulted; **Lobe retirement still the operator's call**). Waterland, Homepage/Skyfall, fleet `.internal` DNS all landed earlier and are stable.
+- **🟢 OTHER SERVICES.** `hrafn` browser-fetch adopted on ana-docker (infra-ops owns uptime; CI now genuinely deploys). speaches ASR live irv-ml1:8204. Open WebUI esh-docker-vm:3211 — **Lobe retirement still the operator's call**. Booth gained kept-board deletion + per-row link pruning. `pfi` gitea org created; claude-bot is an Owner and can create repos self-serve.
 
-- **⏳ OPEN:** the MTP-k3-on-new-build isolating experiment; file the drafted upstream vLLM issue (operator's GitHub identity); Cold-Fusion NVFP4 quants (44 GB) delete/keep; OWUI image-tag drift (`:main` vs pinned v0.11.0); `/tank` DEGRADED **70+ days**; Brokkr duplicate `reranker-a3-bge-v2-m3` alias; **MANY commits unpushed** — push is the operator's call.
+- **⏳ OPEN ELSEWHERE:** MTP-k3 isolating experiment; upstream vLLM issue to file (operator's GitHub identity); Cold-Fusion NVFP4 quants (44 GB) delete/keep; OWUI image-tag drift; `/tank` DEGRADED **70+ days**; Worldtree **#411** debug-room litter; bridge/engine agent-roster drift on both WT instances; brokkr's `gen` vs trained-reward-model bake-off (theirs to initiate). **Working tree is clean and pushed** through `0ad332b`.
 
 ## Recent decisions
 
+- `[2026-08-23]` **Anaheim's IPsec tunnel delivers ~25% of a verified 2 Gbps circuit — not WireGuard, not CPU, not the fibre.** Cipher change proposed and operator-signalled; **execution pending, untracked by operator choice.** → `persistent-memory.d/2026-08-23-anaheim-ipsec-tunnel-ceiling.md`
+- `[2026-08-23]` **selene retired after losing a head-to-head on its own job; `chat-judge` moved to gen, the model name 404s by design.** Also surfaced that **7 aliases share one seat** — cross-checking between them is an echo, which caught a real defect in brokkr's 46k-exposure R47 gate. → `persistent-memory.d/2026-08-23-selene-retired-alias-collision.md`
+- `[2026-08-23]` **hrafn adopted; its CI reported green for its whole life while deploying nothing.** A staging dir inside the rsync target destroyed its own source mid-copy; the deeper fault was verify steps that asserted uptime, never content. → `persistent-memory.d/2026-08-23-hrafn-adopted-ci-frozen-source.md`
+- `[2026-08-23]` **Worldtree b187 shipped; all three instances de-armed from a 69-day-stale `:latest`; Matrix homeserver re-plumbed to personal.** Includes the `:8009`-is-demo port trap that an IP-only fix would have walked into. → `persistent-memory.d/2026-08-23-worldtree-b187-pins-matrix.md`
+- `[2026-08-23]` **Every secret-bearing `.env` on ana-docker tightened to 0600** — eight stacks including vaultwarden and traefik, verified exposed by reading one as `nobody`. → `persistent-memory.d/2026-08-23-ana-docker-env-perms-sweep.md`
+- `[2026-08-23]` **`pfi` gitea org created; claude-bot is an Owner and creates repos self-serve.** Closes the repo-creation half of the credential-migration directive — `vh` is a USER namespace so no service account could ever create there. Repo creation needs `write:user` + `write:repository` + `write:organization`; `POST /users/{u}/tokens` is basic-auth only, so minting needs the account password. Default new repos to `pfi/`. (`vh/eitri-smithy` was its first tenant, then moved.)
+- `[2026-08-23]` **Booth: kept boards are deletable and link rows are prunable.** `release` on a kept card drops the sentinel so the existing × applies; `booth links` / `booth unlink <id|index>` prune one row. Rows are addressed by **content id, never position** — the board is append-only and multi-writer. **Releasing a board RESETS its TTL clock** (unlink bumps the dir mtime), so unkeep-and-wait is a 24h delay, not a delete. (`4be880f`, `0ad332b`)
+
 - `[2026-08-22]` **DFlash2 spec-decode measured on our own stack; `sec` promoted to it.** +18–21% accepted length and +15–18% throughput over MTP k=3, drafter proved model-agnostic across two finetunes to 0.06%, and the k=7 MTP *control* showed deeper MTP is a throughput trap. → `persistent-memory.d/2026-08-22-dflash2-spec-decode.md`
 - `[2026-08-22]` **Quant pipeline shipped a crippled tokenizer for months — fixed at source.** `quant_mixed_nvfp4.py` baked its calibration truncation (`max_length 2048`) into every mixed-NVFP4 build; latent on old transformers, fatal on new. Both live quants corrected, pipeline now saves a source-pristine tokenizer and asserts it. Playbook §3.14. (`0755ba7`)
 - `[2026-08-22]` **`sec` retuned to util 0.52 / 420K after a runtime OOM at 0.55/480K** — `gpu-memory-utilization` is not a hard reservation; activation grows past the dummy-data profile and six vLLM containers share GPU1. Also measured: the KV pool varies ~6.6% between boots, so max-model-len must be sized against the *lower* observation. (`6e82899`)
@@ -256,6 +264,9 @@ _As of 2026-08-22 — three AI seats live on ana-ml2; `sec` is the one that move
 _209 older entries archived to archival-memory.md._
 ## Tried and abandoned
 
+- `[2026-08-23]` **A `HEAD == GITHUB_SHA` assertion in the hrafn CI — added, broke the checkout twice, removed.** It needed the `git` binary (run 9920, exit 127); installing `git` then flipped `actions/checkout@v4` off its **node** implementation onto the git binary, which died on a missing CA bundle (run 9921). A nice-to-have assertion changed the checkout's code path and broke a working pipeline. Removed rather than patched with `ca-certificates` — it guarded a hypothesis that proved wrong. **Do not add `git` to that prereq step.**
+- `[2026-08-23]` **Repointing `selene-1-mini-8b` at gen's endpoint — proposed by me, correctly overruled.** *"never repoint a named model at a different model's endpoint — that is intentionally misleading."* The trap is that it does not feel like deception; it feels like sparing consumers a migration. That framing is the tell. Role aliases move; model names die with the model and 4xx.
+
 - `[2026-08-15]` **Grafted bf16 MTP loads UNINITIALIZED (0% accept) unless `re:^mtp.*` is in the quant-config `ignore`; and W4A16=Marlin (not native FP4) costs ~20% even on decode.** Cost a premature 79 GB delete of a good model (declared desync-dead off the 0%). Lessons: test MTP on bf16 FIRST, isolate before deleting; modelopt 0.43 is dependency-hell for qwen3_5 (list-vs-dict quant_cfg + transformers conflict) — use llm-compressor. Full → `persistent-memory.d/2026-08-15-uncensored-gen-seat.md`
 
 - `[2026-08-03]` **ComfyUI `--enable-triton-backend` on the irv-ml1 A6000 crashes EVERY render — Ampere has no hardware e4m3.** adhoc-agent's operator-approved probe: comfy_kitchen's triton backend has a FUSED int8 matmul that would beat the eager backend's ~1.9x-slower unfused int8 path (21.3s vs 11.2s fp8 on the Moody Krea2 int8 checkpoints). Flipped it (added to `COMFY_CMDLINE_EXTRA`, recreated) → `triton.compiler.errors.CompilationError: ValueError("type fp8e4nv not supported in this architecture. supported: fp8e4b15, fp8e5")` in `comfy_kitchen/backends/triton/quantization.py:145 dequantize_per_tensor_fp8`, failing at **node 5 CLIPTextEncode**. Triton's fp8 dequant kernel targets `fp8e4nv` (Hopper/Ada e4m3); **sm_86 Ampere (A6000) lacks hardware e4m3** → the JIT compile dies. With triton on it grabs the **global** `--fp8_e4m3fn-text-enc` dequant, so every render (fp8 AND int8) dies upstream at the text-encode step — the int8 UNet path never ran, so the convrot-coverage caveat wasn't even the limiter. Reverted cleanly (~15s to healthy, image unchanged `sha256:94afb8ca`, sage intact, prod restored). **The parked cu130 rebuild won't fix it** (e4m3 = hardware format, not CUDA version). **DEFERRED to the Ada refresh** (operator: "ada is coming, we'll optimize then" — Ada sm_89 has native e4m3, so triton's fp8 path should compile there). **Mechanics:** `--enable-triton-backend` is a compose `environment:` var, so toggling it needs `docker compose up -d` (**recreate**), NOT `docker restart` (reuses the baked env, no-ops silently). Full: auto-memory `parked_triton_backend_ampere_fp8`.