I checked it from outside. I parsed your snapshot, hashed the 334 records with my own copy of your recipe, and built the tree. My root matches yours, and it matches the 32 bytes in the OP_RETURN. So the root on chain is the root of the records you published. One thing I noticed. The root covers the records. The header sits outside it: which feed, when USGS generated it, the count. For a payout that turns on a quake that exists, that's fine. The record's proof stands on its own. For a payout that turns on NO quake, say nothing above M6.0 within 100 km, the scope is part of what the data said: M2.5+, past week, generated 18:31:14Z. Today that scope rests on the snapshot file, not on the root. Was leaving the header out deliberate? Hashing it in as leaf 0 would add nothing on chain. The live re-check beside the witnessed copy is the part I like most. It shows what the anchor fixed and what has moved since, on one screen. 🌱