The conclusion table first
Row one asks whether the site answers at all. This is the widest question in the set, it changes slowest, and it can stay true for months after everything useful has already stopped happening there. Row two asks whether orders are being taken, which is a different matter because a storefront can serve pages happily while the buying side sits switched off. Row three asks whether a rebuilt marketplace address is accepted, which decides whether an old file is a transferable list or a stack of re-entry work.
Row four asks whether one specific row still leads somewhere today, and it is the only line this site can answer with its own equipment. It needs a rebuild step and one request per row, which is what the weekly snapshots do, with the values printed on the ledger beside the sample identifiers. The other three rest on public pages, operator notices and third-party records, each read on a stated day. Labelling the source of every row is the point of the table.
The table deliberately keeps four verdicts apart instead of reducing them to one word, because a single word is what produced the confusion in the first place. A host that still serves pages, a purchasing side that stopped, a rebuild that keeps working, and one row whose listing has quietly gone can all be true at once. Collapsing them forces a choice about which fact to drop, and whichever gets dropped is the one somebody needed.
Each row carries a reading, the source behind it, the date it was read, and a plain statement of what it does not cover. A row without a date is a rumour with a table around it, which is why the checking method page insists on dating every observation. Period 2026-W39 was read on 2026-09-29 and that date governs the whole table below it. No line here was written from memory.
Two of the four rows can be partly re-checked by a reader in an evening without an account anywhere. Open a public page and see whether it serves, then rebuild one identifier from your own file and see whether the marketplace answers. The middle two cannot be re-checked from outside at all, and pretending otherwise is how people end up trusting a chat message over a record. The table shows which parts are yours to test.
A reader can treat the four rows as a short checklist and stop guessing after the second line. Opening a public page settles row one in seconds. A dated notice settles row two, or leaves it unanswered. Row three is a field on the ledger with a date attached. Row four is the only line that costs work per row, and a reader comparing their own file against a measured window already has the shape of it.
Why each row reads that way
Row one reads as down, and the path there is recorded rather than remembered. A notice on 2026-06-01 described security and compliance work and mentioned disrupted addresses, slower pages and cancelled orders. Purchasing and redemption stopped on 2026-06-12 and stayed stopped through the rest of June, with warehouse and parcel work continuing under the same notice. A step-by-step return planned for 2026-07-01 did not bring ordering back, a community representative said as much around 2026-07-10 without offering a new date, and by August third-party records showed the storefront offline.
Row two reads as no, and the honest answer has to come from the operator. Purchasing has not been described as restored by anyone speaking for the platform, and no new date has been published since the July statement. This site has no account access and holds no order data, so nothing here can confirm or contradict what a particular account still shows. The dated fact is narrower: the public notices stopped promising a return and then stopped altogether. Row two stays an operator question.
Row three reads as still open in the record and untestable in practice, which is an awkward verdict worth keeping. The platform ledger tracks twenty-two objects and records whether each accepts a rebuilt source address, and those fields are dated like everything else; for a storefront that has stopped taking orders, the field cannot be exercised by anybody at all. It matters for the move rather than for the old platform. Eleven ledger objects also have an entry in the public directory, which is where a reader looks to see who can still say yes.
One rule covers all five labels the record uses. A status describes an address and a day, never a product and never a level of availability. Resolves means a rebuilt source address answered in that window; moved means it answered somewhere else; gone means the listing came back as not found; shielded means a wall blocked a clean reading; unrecorded means the check never finished. A blank cell is the last of those rather than a verdict, and it says the test did not complete.
Row four reads as resolves across the whole period, and this is the row built from our own checks. Ninety-three addresses were rebuilt and requested in the 2026-W39 window, every one of them returned an answer, and not one row was recorded as gone. The thirty-one wrapped rows collected from public sheet pages all failed at the network layer in that same window, so the split sits between the wrapper and the listing. Row four answers a narrow question, and it answers it well.
The conditions that would flip it
Row one flips when the host answers reliably again and a reachable storefront follows, which takes more than one response. It would also flip the other way if a notice appeared describing a permanent shutdown, because a dated statement from the operator outranks an absence of pages. Neither has happened, so the reading rests on the June notices, the July statement and the August third-party record. One successful request would not move it, since that proves a host is up rather than a platform is back.
Row two flips on an operator announcement that ordering has resumed for accounts generally, not on a rumour about a soft reopening. The condition is specific and it is why this row is watched separately: the words would have to come from the platform or from somebody speaking for it, and they would have to carry a date. A partial return restricted to existing balances would be a different reading. Third-party records cannot flip this one, however plausible they sound.
Row three flips either way and is the cheapest row to re-read. It changes when the ledger records a different answer for a platform page, and it is checked against public pages rather than against an account. A platform that starts refusing rebuilt addresses would turn every old sheet into manual work, and one that begins accepting them is the most useful change available to anybody holding a file. A reader can partly verify it alone.
Row four flips when a rebuilt address stops answering, or starts landing somewhere unexpected, which is what repeated checking is for. One vanished listing would move one row and leave the other ninety-two where they were, and that is worth seeing written down rather than assumed. Watch this row on the snapshot schedule instead of on demand, and read the date at the top first.
One habit protects all four rows and costs nothing: read the date before the verdict. A reading from an old window is not wrong, it is expired, and treating an expired reading as current is what turns a record into an argument. If the date sits before your own copy of the file, check again rather than debating what the number meant. The date decides how much of the table still applies.
A last condition bounds all four rather than flipping any of them. This site holds a single window of its own readings, period 2026-W39, and the sample behind it was drawn from public pages rather than assigned at random, with a further hundred and thirteen addresses left out when the host stopped accepting connections. Any later reading gets compared against that baseline. Until a second window lands, the table claims nothing about a direction of travel.
If this site stops publishing, the four rows become fixed rather than useless. The last window on file would stay a dated comparison for as long as anybody keeps it, and the rows a reader can test alone would keep working without help from us. What would be lost is the second period and everything it makes possible. Nothing in the table depends on this site reading it next.