Mainnet relay testing for DOG Mode: a kit, and an offer to run it for you

[dogofbitcoin]

Both mainnet tests from the 2026-09-06 list are done: the 1 sat output (#3, comment 5666936067) and the 3,847,522 WU transaction carrying a real ordinals envelope (#3, comment 5722763923). We ran them with a small kit we wrote for exactly this kind of test, and we would like it to be useful for more than those two.

What the kit does. One Python file on a laptop, with a throwaway key that never sits on a server. For a given transaction shape it:

• builds and signs the transaction locally, then asks each DOG Mode node testmempoolaccept before anything is sent;
• sends it through one node and times its arrival at the other from each node's own getmempoolentry, with mempool.space as the Core-policy control;
• pulls the raw transaction back out of each mempool with getrawtransaction and checks the payload byte for byte against what was sent;
• takes the coin back by key path through a Core node, so an oversized transaction is never mined and its fee is never paid, then sweeps what is left home.

Both runs together cost 1,268 sats in fees, start to finish. Before the big run, the nodes caught two bugs in the kit that its own offline checks had passed: embit's sighash_taproot defaults ext_flag to 0 for a script path spend, and a 7.7 MB hex does not fit in one execve argument. Both are fixed and pinned by tests, and the kit now refuses to sign when two independent builds of the signature message disagree.

The offer. If you are working on relay behavior in DOG Mode (the weight limit, dust, datacarrier, the reserved dog slots, anything that is policy), tell us the commit and the shape you want tested. We run it on mainnet across two dog nodes, a full node built from source and the pruned Umbrel package behind a VPN, and post the readings here with every txid so anyone can check them.

The write up of the big run: https://seed.dogofbitcoin.org/4meg. The script that verifies it against the chain, with the video, so anyone can run it: https://github.com/dogofbitcoin/dogmode-relay-tests

One reading we already hold that bears on a release: our full node still fills 0 of its 4 reserved dog slots and the Umbrel fills 1. We will measure both again 24 hours after a release tag.