docs: matrix → 0.15.17 published; fix flagship RAM/wall-time to match REPORT.md

- README version matrix: 0.16.6-sqd.1 = latest, 0.15.17-sqd.1 = published (both live on npm); versions.json published:true.
- harness/euler-multichain/README.md "Hardware & wall-time" contradicted REPORT.md — it claimed more RAM
  gives better numbers. Corrected to match the measured run: single-thread indexing is the ceiling, the
  16 GB/2-core cap was faster than 32 GB (44m 55s vs 67m 10s), and the lever is sharding, not hardware.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: dzhelezov

README.md

@@ -77,8 +77,8 @@ Full adoption path — check, swap, run, validate, roll back — in [`MIGRATION.
 
 | Ponder | `@subsquid/ponder` | Status |
 |---|---|---|
-| 0.16.6 | `0.16.6-sqd.1` | current |
-| 0.15.17 | `0.15.17-sqd.1` | planned |
+| 0.16.6 | `0.16.6-sqd.1` | **latest** |
+| 0.15.17 | `0.15.17-sqd.1` | published |
 
 `npm install @subsquid/ponder` installs the latest; pin `@X.Y.Z-sqd.<rev>` to match your Ponder version. How versions work: [`PUBLISHING.md`](PUBLISHING.md).
 

harness/euler-multichain/README.md

@@ -49,15 +49,12 @@ live. Pair with Postgres `pg_stat_activity` wait-events to see whether the DB is
 CPU-bound, or *waiting on the client* (i.e. upstream decode is the limit).
 
 ### Hardware & wall-time
-The Portal is **not** the bottleneck here — local decode + DB write + the in-memory fetch buffer are,
-and that envelope scales with **RAM**. On a memory-constrained box (the captured run: 16 GB) the
-24-event superset must use conservative, uniform chunks (`PORTAL_CHUNK_FIXED=1`,
-`PORTAL_CHUNK_BLOCKS=300000`) to bound the buffer, which trades **more HTTP round-trips** for a fit —
-*same Portal CU* (CU is billed per data-chunk, not per request), just added latency on high-block
-chains. On a larger-RAM machine you instead keep the default **density-scaled chunks** (arbitrum →
-multi-million-block chunks) and a bigger heap, cutting round-trips dramatically → **materially better
-wall-time**. So the reported wall-time is a **conservative floor on modest hardware, not the fork's
-ceiling** — more RAM gives strictly better numbers.
-
-See **[REPORT.md](./REPORT.md)** for a captured full run — throughput, per-chain reach, the
-saturation analysis, and the hardware caveat above applied to the actual wall-time.
+The Portal is **not** the bottleneck here — Ponder indexes on a single thread, so local decode and DB
+write are, and that ceiling does not move with RAM or cores. The captured run capped the indexer at
+**16 GB / 2 cores** on an otherwise large box, and that modest cap was **faster** than a 32 GB
+configuration (**44m 55s** vs 67m 10s): extra heap and a deeper read-ahead buffer just sit idle when a
+single CPU thread is draining them. The lever past this ceiling is **sharding** — splitting chains
+across processes — not bigger hardware.
+
+See **[REPORT.md](./REPORT.md)** for the captured full run — throughput, per-chain reach, the
+saturation analysis, and the 16 GB vs 32 GB comparison.

versions.json

@@ -10,20 +10,23 @@
       "wiring": "wiring/0.16.6.patch",
       "rev": 1,
       "status": "verified",
-      "published": false,
+      "published": true,
       "note": "bare 0.16.6 seed-publish (2026-06-30) is BROKEN (workspace: deps) — unpublish it. 0.16.6-sqd.1 is the publish candidate: byte-identical vs RPC across logs/tx/receipts/traces (harness/diff) + C1/C4/C6/C7/C8 fixes + 17 unit tests. Publish with --tag latest."
     },
     {
       "ponder": "0.15.17",
       "wiring": "wiring/0.15.17.patch",
       "rev": 1,
       "status": "verified",
-      "published": false,
+      "published": true,
       "note": "Seam identical to 0.16.6 — the 0.16.6 wiring patch applies cleanly, builds, and passes all 53 tests on ponder@0.15.17."
     }
   ],
   "compat": {
-    "tested": ["0.16.6", "0.15.17"],
+    "tested": [
+      "0.16.6",
+      "0.15.17"
+    ],
     "seamStableRange": "0.15.17 - 0.16.6 (verified identical by inspection; CI matrix promotes to 'tested' as patches land)",
     "howToAdd": "see PUBLISHING.md — scripts/sync-upstream.sh <ver> --test, then add the row here"
   }