The Liquid Network is producing blocks again, and the chain it is producing them on is not the one it stopped on. At 06:11:10 UTC on Friday 11 September, Blockstream’s own Liquid explorer API returned a tip height of 4,052,304. Walk that chain backwards and the seam is unmistakable: block 4,050,336 is timestamped 21:05:10 UTC on Wednesday 9 September and carries four transactions including its coinbase, and it sits directly on top of block 4,050,335, which is timestamped 13:52:10 UTC on Sunday 6 September. Three days, seven hours and thirteen minutes separate two consecutive blocks on a chain that normally produces one a minute. Block 4,050,336 is the height that the Liquid Federation’s incident report of 8 September names as the block in which the Elements bug was exploited to create roughly 4,000 unbacked L-BTC. The block that now occupies that height is dated three days after the exploit. Whatever block was there on Sunday is no longer served.

That matters for a reason the desk can document from its own records. On Thursday morning, at 06:17 UTC, this desk queried the same API and recorded a tip of 4,051,232, the height NFTevening had reported on Tuesday with a timestamp of 04:49 UTC on Monday 7 September — the point at which the federation stopped signing blocks after the exploit. On Friday morning the block at height 4,051,232 is timestamped 12:19:10 UTC on Thursday 10 September. A block’s timestamp is part of its header, so a different timestamp at the same height is a different block. The 897 heights from 4,050,336 to 4,051,232 inclusive, which on Monday held roughly fifteen hours of post-exploit history, have been re-filled with blocks produced on Wednesday night and Thursday. This is what the federation’s three-stage plan — resume blocks without pegs, “replay transactions verified as valid,” then reopen the peg — looks like when it is read off the chain rather than off a status page.

What the replay looks like on the chain

The first blocks of the new chain are the replay. Block 4,050,336 (21:05:10 UTC) carried 4 transactions (block counts in this article include the coinbase; the 490 figure below excludes it); block 4,050,337, six minutes later at 21:11:10, carried 329; block 4,050,338 at 21:16:10 carried 83; 4,050,339 at 21:17:10 carried 40; 4,050,341 carried 34. From 21:18 UTC the cadence settled to one block a minute and the blocks were, with a handful of exceptions, coinbase-only. Across the first hundred heights (4,050,336 to 4,050,435) eight blocks carried more than the coinbase, for 490 non-coinbase transactions in total, the last of them three transactions in block 4,050,427 at 22:51:10 UTC. By Friday 06:10 UTC the most recent ten blocks carried between one and three transactions each. The desk cannot tell from block headers which of the 490 were replays of transactions from the discarded window and which were new; the federation’s Thursday status, posted at 12:26 UTC and dated 10:00 UTC, said block production had resumed “without transactions” while the network was monitored, which is a description of the empty blocks, not of the 329-transaction one that preceded the statement by fifteen hours.

One timestamp in the incident report does not survive contact with the chain. The federation wrote that the exploit took place “on September 6, 2026 at 15:53:10 UTC (Liquid block 4,050,336).” SideSwap’s own statement of 9 September, as rendered by The Crypto Times, puts the inflation transaction at 13:53 UTC at the same height. Block 4,050,335 is stamped 13:52:10 UTC; on a one-minute chain the block after it was produced at 13:53:10 UTC. SideSwap’s clock is the one that agrees with the chain; the federation’s figure is exactly two hours later, which is what 13:53 UTC reads as on a machine set to Central European Summer Time. It is a small thing, and the desk flags it because a reader reconstructing the timeline from the federation’s report alone would place the exploit two hours after SideSwap’s test peg-out rather than seven minutes before it; the desk cannot tell from the outside whether the offset applies to the report’s other times.

Two new messages from Blockstream, signed and sealed

On the Bitcoin chain the conversation resumed, but only on one side. Since the whitehats’ 588-character demand confirmed at 11:45:49 UTC on Wednesday, 25 transactions have landed at their address, bc1ql4mfu6…, and two of them are Blockstream’s. The first, 32bd329b…, confirmed in block 966,296 at 02:18:58 UTC on Thursday 10 September; the second, 6819a709…, in block 966,410 at 21:48:17 UTC the same day. Each pays the whitehat address 1,000 satoshis and carries an OP_RETURN of about five and three kilobytes respectively. The desk extracted both payloads and ran them through gpg against the key Blockstream publishes at blockstream.com/pgp.txt. Both return “Good signature” from RSA key 1176 542D A98E 71E1 3372 2EF7 4AC8 CC88 6844 A2D6, “Blockstream Security Reporting <security@blockstream.com>,” the same key that signed the “safe to return the funds” message on Monday. The signature timestamps are 01:52:37 UTC and 21:32:12 UTC — 26 and 16 minutes before each transaction confirmed.

What the signatures cover is not readable. The signed body of each message is a single block of base64 text whose first four decoded bytes are BIE1, the header that Electrum-compatible wallets prepend to a message encrypted to a Bitcoin public key. Blockstream, in other words, signed its messages so that anyone can verify who sent them and encrypted them so that only the holder of the whitehats’ key can read them. That is a change of posture on both sides. Blockstream’s four earlier on-chain messages were plaintext (one unsigned, three signed). The whitehats, in their 8 September message, had written “All messages will be in plaintext,” and their 9 September demand was plaintext. The two Thursday messages were sent from addresses the desk has not seen in this conversation before, bc1qt367… and bc1qys9c…, rather than the bc1qn8mg… address Blockstream used from Sunday to Monday; the signatures, not the addresses, are what tie them to Blockstream.

The whitehats have not answered. As of 06:12 UTC on Friday, the address holds 598.50048115 BTC across 296 confirmed transactions — 8,046 satoshis more than on Thursday, all of it dust attached to incoming messages, Blockstream’s two 1,000-satoshi payments included — and has not spent an output since the 3,400 BTC return at 16:09:25 UTC on Monday. A further 34 unconfirmed dust transactions sat in the mempool addressed to it. The other 23 confirmed arrivals since the demand are third parties: a self-described “pleb negotiator” at 22:24:47 UTC on Wednesday offering to mediate and floating a compromise of “50–100 BTC out of Blockstream’s own pocket”; a message at 11:38:13 UTC Thursday reading, in full, “Sequentia ignores, fixes, pays zero. Not a whitehat target. Do not refund.”; six messages between 07:49 and 21:48 UTC signed “the onchain 10btc beggar”; and an advertisement for a list of mixers. At Thursday’s Bitstamp close of $76,528.75 the 598.5 BTC is worth $45.8 million, the 3,400 returned is worth $260.2 million, and the 399.6 BTC that would satisfy a 10% demand on 3,996 is worth $30.6 million.

“Will be covered” — and what that does and does not say

Adam Back, Blockstream’s chief executive, wrote on X at 11:02 UTC on Thursday that “network resumption” was coming and that the L-BTC-to-BTC peg would be covered one-for-one, adding “Do not panic sell OTC,” according to Bitcoin.com News and The Crypto Times, both rendering the same post; at 13:13 UTC he followed with “and @liquid_btc – we’re baaack!” The federation’s status said the same in institutional language: “restoring the BTC/LBTC reserve is in progress.” Neither says where the 598.5 BTC comes from if the whitehats keep it, and that is the distinction Field Guide #45 drew on Thursday: a federated peg’s shortfall is a queue, not a haircut, until someone names the source of the cover. Bitcoin.com reported the point the same way: Back “did not explain how the near-600 BTC hole will be covered when resuming peg-out operations.” Marker Y1 — a statement by 30 September of full backing from non-whitehat sources, or a bounty paid or agreed — therefore stays open. “Covered” is a promise about the outcome; Y1 asks for the source.

SideSwap’s statement closes one small loop the desk left open on Thursday. It lists a 2.5 L-BTC test peg-out at 14:00 UTC on Sunday paid as 2.49749857 BTC, the 4,000 L-BTC order at 14:05 burned at 14:06, the federation’s 3,996.02 BTC payout at 14:28, its own forwarding of 3,995.99999857 BTC in the same block — and says SideSwap later returned its 0.1% peg-out fee, about 4 BTC, to the federation. On Thursday this desk reported a separate, unexplained 4.0 BTC arriving in the federation’s peg address at 08:13 UTC on Wednesday from three unrelated addresses. The desk cannot prove the two are the same coins, but the amount, the direction and the timing are consistent with SideSwap’s description, and it is the only candidate on offer. SideSwap also accepted two operational failures in its own words: it kept its peg-out authorisation key online, so payouts could land in the same Bitcoin block as the order, and it applied no size, rate, velocity or wallet-history checks to peg-out orders. The Elements bug created unbacked coins; the hot key and the missing limits are what turned them into 3,996 BTC on the Bitcoin chain.

What is live, what is not, and what the markers say

Live: block production, since 21:05:10 UTC on 9 September, at one block a minute, with transactions in some blocks. Not live: peg-ins, peg-outs, PAK-authorised or otherwise, and, by SideSwap’s account, its instant swaps. The federation peg address held 3,601.47203772 BTC at 06:12 UTC Friday, unchanged from Thursday to within the dust. Against the 4,205 BTC the incident report gives as the pre-exploit reserve, that is 85.6%. Nothing on either chain has changed that ratio since Monday afternoon.

The markers grade as follows. T1 — at least 3,500 BTC moved to the peg address by 00:00 UTC on Monday 14 September, the end of Sunday — is failing by exactly 100 BTC (3,400 returned by the whitehats against a 3,500 bar) with two days left, and the only party that can change that has been silent since Wednesday morning. V1 — a Liquid restart and a federation peg-out by 30 September — is half met: blocks yes, peg no; it stays open. Y1 stays open for the reason above. The desk adds one marker from today’s reporting. AA1: the whitehat address bc1ql4mfu6… spends at least one output before 00:00 UTC on Thursday 17 September. Blockstream has now written twice, privately, in under twenty hours. If those messages contain an offer, the response that counts is a transaction, not a message; if they do not, a week of silence from the coin-holder is itself a reading. Either way the chain will grade it.

For the vocabulary — federation, PAK, why a consensus bug is not a key theft — see Field Guide #44; for how to read the 10% demand itself, Field Guide #45; and for how to verify a rollback like this one with nothing but a block explorer, today’s Field Guide #46. The desk’s Monday and Thursday reporting on the incident is here and here.

Sources: blockstream.info/liquid API for block heights, hashes, timestamps and transaction counts, pulled 06:10–06:17 UTC 11 September; mempool.space for every Bitcoin transaction, timestamp, fee and balance, pulled 06:10–06:12 UTC; blockstream.com/pgp.txt and GnuPG for the two signature verifications; the Liquid Federation’s incident report (X, 8 September 19:10 UTC) and resumption status (X, 10 September 12:26 UTC) as rendered by The Crypto Times and Bitcoin.com News; SideSwap’s 9 September statement as rendered by The Crypto Times; Adam Back’s posts of 10 September as rendered by Bitcoin.com News and The Crypto Times. Quotations from X posts are the renderings of the outlets named. Prices are Bitstamp’s 10 September close.

Method: prices, funding, open interest, basis and on-chain figures in this article are pulled directly by Bitcoin Mastery at the timestamp stated — Bitstamp BTC/USD daily candles for closes, Binance BTCUSDT spot and USDT-margined perpetual for intraday, open interest, funding and account ratios, Binance COIN-M quarterly contracts for basis, mempool.space for difficulty, hashrate, pool shares, fees and individual Bitcoin transactions, blockstream.info’s Liquid API for sidechain block heights, hashes, timestamps and transaction counts, alternative.me for the Fear & Greed series, Farside Investors’ table for ETF flows and US Treasury CMT par yields for rates. Where a third-party figure is cited we name the source and its date; where two sources disagree we print both. Every streak or extreme figure is published with the first date of its series in the same sentence.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Cryptocurrencies are volatile and you can lose money. Nothing here is a recommendation to buy or sell any security, digital asset or exchange-traded fund, including MSTR, L-BTC or HYPE. Do your own research and consult a licensed financial advisor before making investment decisions.