Thank you for checking the chain side. The one byte. A signature is 71, 72 or 73 bytes, depending on the numbers that come out of each signing. My replay can't use the anchor key, so it signs with a throwaway key, and the length moves by a byte from run to run. I ran it again: the first transaction came out at 13,621 bytes, exactly 84829d3b, and the second came out one byte short of Start. Outputs, fee and rate are the same every time. The replay is yours to run. Two files: anchor-wx-site-70dff36.src.mjs is the anchor code exactly as it was at 18:25 on Nov 16. replay-84829d3b.mjs runs that code with @bsv/sdk 2.7.1 against dc0cfbd8…:2 and 0dacb3c4…:0 from chain, then compares what it builds with 84829d3b and Start. The broadcast is replaced by a stand-in, so nothing is sent. The commit times stay mine. Some of what you couldn't check is on chain. Every transaction below carries an OP_RETURN labelled bsvkey.proofmap.test. They're signed by 1ErDfgzGWe6kDSHWWZPUSpVRRvfQo7rDdZ, not the line key, so under line/1 none of them count. Test line, second run: e1f8b0f4ea96e85fc516b4dea9ef27c14c2b5f2027304a3ffd04ed5092f0acf4 spends 116140e0…:1. 41f1ccfcadd93af2ad0907161556b3fd20523a51fb4d5a7bcba40ac7bb345e6f spends e1f8b0f4…:2. be184a6288373a55b081832c577f7fc51ae222d6c88e06a5dcce5366ce1bf704 spends 41f1ccfc…:2. The code was told the tip was e1f8b0f4…:2, which 41f1ccfc had already spent. It found 41f1ccfc while that was still unmined, in 2 seconds. 185ae98e54b88ff4d4b8a52d1425ca077370558fdf565d1a92c416a0aad4707d spends be184a62…:2 plus 7b0448…:1, because its data cost more than the tip held. The tip stays input 0, and one output pays the key. Test line, first run (same shape; there the stale-tip step waited for a block): 13e9be21…, 3443ab5d…, 545565b6…, 5e453c0b…. The four fee tests, all mined in block 976,411: 1 sat/byte: b0a4281c6cb33ec8638d66308655a471062554c577363f067d144c48b9d12c26 0.5: 40c8e9d065768234c0ddf74a867f8ce23749061906b18ef033fe9a0e22253bc9 0.25: baf0f21a4b533ed4abc139ec0189070d2f43662e6d0d3e8ec5752dc2bdb82172 0.1: a7f161eeae1fdcf3983416499f51dc1343a18b36072c82f00525ba9b5eb72b85 On your last point: yes. A refused anchor costs a retry, and a dropped line output costs the line. The code now treats the line output as the one thing it never gives up. The same fee fix is now on the other BSVKey sites. The two open questions: settled, and on chain. A later rule may change only how anchors after it count. The core is fixed. Every anchor is judged by the rule in force where it stands in the line. Nothing already in the line is judged again. line/2 is in the line now: 082e865c727a166766655d9c00e5025c38b1ad0c02c59487422e126fd36b9198, block 976,431, 04:28 UTC on Nov 21. It spends c93173904c1caf4c…:2, the tip you checked, and adds one coin for the fee. Output 1 is the data, a snapshot of 346 quakes, and it rebuilds. Output 2 carries line/2 in full, in the same four pushes as line/1: bsvkey.proofmap, bsvkey.proofmap.line/2, text/plain; charset=utf-8, the text. It is 2,701 bytes, sha256 057662776389ebe43f262084868f80be79f3cf86b84a7611be5dd475804807df. Output 3 is the line output, 35,309 sats, and it is unspent: the tip. The fee is 4,691 sats for 18,759 bytes, 0.250 sat/byte. The line is 18 anchors. Because the key signed this anchor, it has now signed a rule that binds it. line/1 never had that, since it was published from the recipe's wallet. Here is the text, so you can check it against output 2: