Candidate: f3a14e6528415c7c98da7865eb6492becb8c1c61, clean main on Studio and
/root/xln on xln.finance. This records actual Anvil testnet deployment, not
mainnet activation, Sepolia financial exit proof or completion of the release bar.
The previous primary Anvil checkpoint failed to parse and RPC8545 was unavailable; old hubs repeatedly failed restore. The current canonical deployment artifact was built from this SHA and installed with all1322 file hashes verified. Both Anvil chains31337/31338, three hubs, MM, custody and one watchtower are now ready.
https://xln.finance:8/8,26.867s; /app redirects to React /ui/, fresh
wallet and Account worker initialize, relay receives12 frames, deep route
reload/unlock works, pinned compiler preview works with financial admission
unavailable, and graph observer remains available. TLS verification enabled.
Page errors, console warnings/errors and HTTP failures:0.714c22744cbc9fbbb626e192ba7ee3e878ed1513e16846f381938b5b9a01c54a;
Account worker 5eb1a38aa8d02a663289ba35ba286e2d8067d373bb7bc31e80a61887ad15a945.Only existing PM2 owners anvil, anvil2, xln-server, xln-watchtower were
stopped. Zero handles were verified before deletion of the exact obsolete XLN
stores recorded in reset-result.json. Existing runtime seeds and watchtower key
retained identical hashes and0600 modes before and after startup. Unrelated PM2
owners retained PID/status. No mainnet/Sepolia transaction, credential rotation,
firewall/auth/TLS change or unrelated store deletion occurred.
Remote source used the canonical clean/ancestry/hidden-index guarded fast-forward, not reset/clean. The old generated web archive remains outside source. Nginx gained only the canonical wallet include; removing that exact include reproduces the previous site bytes. Configuration validation and reload succeeded.
The persisted incident journal was NOT deleted. Its16,728,524-byte pre-resolution
copy has SHA256 227477bc0e0e7944786a2d757c4b5bdf1709b3f458f4abaeab7689fa21a7bae5
and is preserved on both hosts outside source. All17 open incidents predated the
new reset/start. incident-resolution.json records each fingerprint, prior count,
last observation and concrete resolution evidence. The operator API resolved each
only after checking its record had not changed; new occurrences reopen it as unread.
This does not claim obsolete oversized frames can now be restored.
The first bundled stand tool deadlocked during module initialization, before any service action. Its exact one-member group was stopped; byte-identical canonical source modules then supplied the stand. An old checkout’s missing stand-ignore rule caused a dirty-source guard refusal; the exact current rule was temporarily added to local exclude and original exclude bytes restored after fast-forward. No guard was bypassed. The first45s bootstrap observation timed out; managed services continued and the next bounded observation found all8 stages ready. Health’s canonical reset start/end timestamps measure114.598s, not the16ms later observation latency.
The public browser is a gateway/runtime proof, not a payment/exit or soak proof.
Aggregate release/full browser, public Sepolia financial proofs, rollback,
uninterrupted one-hour soak and MacBook catch-up remain outstanding. Command exits,
including failures, scripts and remote snapshots are in command-evidence.tar.gz.