The claim you see most
The claim arrives in the same words almost every time: a row marked as resolving means the listing behind it is still there to order. That reading is understandable, because the label sits in a column that looks like a verdict and the word itself sounds like a green light. What the label actually answers is narrower, and the gap between the two is where most sheet arguments begin. The holder of a sheet rarely reads the column header, and almost never reads the method behind it.
The column this ledger publishes carries one value per row, read on one day, from one address. Ninety-three records went into the 2026-W39 window and every rebuilt source address answered, which produces a wall of identical values. A reader scanning that wall fairly concludes that nothing is broken. Nothing in the address layer was broken, and that is a much smaller statement than the one being made. That uniformity makes the column useful in a first period and useless as a forecast, because a value everybody shares cannot separate one row from another.
Stock, seller behaviour and price sit outside the question entirely. An address can answer while the seller has stopped shipping, while the size you wanted is gone, or while the price moved twice since the sheet was written. None of those conditions were measured on 2026-09-29, so none of them can be read off the column, in either direction. Anyone who wants to know about a seller has to look at the seller, and that work does not get shorter because a link column came back clean.
The cost of the misreading is practical rather than philosophical. Somebody treats a sheet as a shopping list, pastes thirty addresses into an order box in one sitting, and discovers at the end that the addresses opened and the items did not match expectations. Rebuilding and checking the addresses was never the slow part. Deciding what you still want, item by item, is the slow part. Addresses are the part that can be settled once and written down, which is exactly why they get asked to carry conclusions they cannot support.
A counterexample
Take the clearest counterexample in this period. Thirty-one wrapped addresses copied from public sheet pages failed at the network layer, every single one of them, and the thirty-one source addresses they encode all answered normally. A status read off the wrapper says the row is dead. A status read off the source says the listing is there. Both statements are true of the same thirty-one rows in the same week, and only one of them is about the item.
The lesson is about which layer the label describes, and it repeats in smaller ways. During the same window, one hundred and thirteen directory addresses were set aside because the host stopped accepting connections. They were excluded rather than written down as unknown, which is a choice about honesty and also a choice about reading: a blank cell in a sheet belonging to somebody else usually means nobody looked. Where a cell was never filled in, the absence is evidence about whoever maintained the sheet, not about the listing itself.
Then there is the wall. A check that comes back as a verification step or a login prompt records that a reliable reading was not obtained; it does not record that the listing was removed. Readers who have watched a platform go down tend to file walls under bad news, and that habit turns a missing measurement into a finding it never was. The label says shielded for a reason, and the reason is caution. Reading a wall as a removal is the most common upgrade of a non-result into a result.
Each of the five labels answers exactly one question: did the rebuilt source address respond, and if so, where did it land. MOVED means it responded somewhere unexpected. GONE means the host returned a page saying the listing is not there. SHIELDED means nobody got a clear answer. UNRECORDED means the check never produced a usable reading. Four different situations, four different next steps, and collapsing them into one word, failed, throws away the only part of the reading that says what to do next.
Why the claim breaks
The claim breaks first on its subject. A status is a property of one address on one day, and the sentence people actually want is a property of an item, a seller and a warehouse at once. Only the first of those four things was ever put to a test, so the conclusion has three untested parts riding along inside it, and those parts are the ones that decide whether an order goes well. They are also the three a buyer cares about on the day a parcel is packed.
It breaks second on time. Every record in this ledger carries the day it was read, and two rows can only be compared when both carry a day. A sheet assembled over several months, with rows typed at different moments and dated only by position in the file, cannot support a statement about the present. The weekly snapshots keep a period label for exactly this reason. A date is not decoration on a row; it is the difference between a measurement and a memory.
It breaks third on the counting rule. Ninety-three records went into the period and one hundred and thirteen candidates did not, because the host stopped taking connections. Both numbers describe the same stretch of work. Anyone quoting one without the other is describing a rule rather than a result, and the rule is always worth asking about before the number is repeated. The gap between those two figures is the honest size of what this ledger does not know about the same week.
It breaks fourth on the ladder. Readers rank the labels by how bad they sound: resolving is good, moved is mildly bad, gone is final, shielded is somewhere in between. That ranking does not survive contact with a single case. A moved address may land somewhere unexpected without the item having moved at all, and a gone page records what the host returned on one day rather than a seller who announced a departure. A ladder feels tidy and it hides the fact that each label describes a different kind of uncertainty.
A wording that holds
The wording that holds is dull on purpose. The rebuilt source address answered on 2026-09-29, and nothing about the seller, the stock or the price was measured. That sentence covers what the check produced and states the boundary in the same breath, which means it cannot be quoted later as a claim it was never meant to make. It also survives being pasted into a message, which longer explanations do not.
Put it in two columns rather than one. The first records what answered, with the day. The second records what you decided from it, in your own words, and it is allowed to be wrong later because it is labelled as a decision. Mixing the two produces the argument this article is about, where a check result is quoted as proof of an item being fine. Two columns also make the sheet easier to hand over, because the next reader can see which half is evidence and which half is opinion.
For rows nobody has touched, the honest cell reads not checked, with the reason if you know it. The period table does exactly that with one hundred and thirteen candidates: they are absent from the counts rather than folded into a failure total. A sheet that writes not checked is doing more work than a sheet that writes gone, because it tells the next reader where to spend an afternoon. A failure total invites an argument about the size of the failure; a not-checked cell invites somebody to go and check it.
One habit keeps the wording intact after the sheet changes hands. Whenever a value is copied from one file to another, copy the day with it, and whenever a day is missing, write that the day is missing. The checking method page sets out the same discipline for the public sample, and it is the reason a row can be re-read in a month without anybody having to trust a memory. Rows outlive the people who typed them, and a row carrying its own date needs no storyteller.