Determination, disruption & cessation

Status: published 2026-08-20 as oracle build item O1.8, which states that the on-chain feed must not ship before this document exists. Extends, and does not replace: cessation.md and restatement.md. Where this document is silent, those govern. Where it adds an obligation, the addition applies to licensed settlement references only (settlement-reference.md).

1 · Why a second document

cessation.md is complete and correct for the reader it was written for. It says what happens when a benchmark stops, gives thirty days' notice "so anyone referencing it can transition", and guarantees the history stays published. For a journalist with a chart, that is everything they need.

A contract needs four things that page does not give: who determines a final value and by when; what the fallback reference is; what happens between notice and stop; and what a consumer is owed when a value they already settled against is later corrected.

And it needs one distinction that page does not draw at all, which is §2.

2 · A refusal is not a cessation — and conflating them is the expensive mistake

This is the most important section here.

A contract that treats an insufficient_data day as a termination or force-majeure event will terminate on a Tuesday, for a reason the index considers normal operation. WETX has already produced five such Tuesdays in a row. Any contract referencing a W.E.T. benchmark must handle the two separately and must not define cessation in terms of "no value published".

The signal that distinguishes them is published and machine-readable. A refusal carries status: insufficient_data with a class and a rule. A cessation is a published notice. Neither requires inference, and a consumer that infers cessation from silence will be wrong.

3 · Determination

Who: the administrator, applying the published methodology as it stood on the settlement date, consulting the independent committee member on any determination departing from the ordinary settlement path.

What the determined value IS: the final settlement row for the date — the row as it stands once the settlement date's UTC day has ended. Before that the row is provisional and may be replaced by a same-day re-run; the two states are distinguished on the record itself and a contract must say which it references.

Timetable: five working days for a determination request; one working day where the answer is that the index refused.

What is never done: a value is not determined by applying a later methodology to an earlier date, by interpolation, by estimation, or by taking a value from another source. If the published methodology does not produce a number for that date, the determination is that there is no number — which is an answer, not a failure to answer.

4 · Disruption — the fallback ladder

For a licensed reference, on a day with no usable final value. W.E.T. supplies none of these fallbacks; the licensee's contract must choose one in advance (§3(2) of the licence). This ladder exists so the choice is made from a named set rather than improvised.

The gap this table closes: cessation cannot occur before day 20, so a contract whose only fallback is "on cessation, do X" has nineteen days of undefined behaviour — which is longer than most contracts referencing a daily benchmark last.

Staleness is not refusal. A refusal is a published statement that the index declined; staleness is the absence of any record, which may mean the publisher is down. They require opposite responses: a refusal is information, and staleness is an escalation. A consumer that cannot tell them apart has already lost the ability to respond correctly, which is why refusals are published and signed exactly as prints are.

5 · Restatement after settlement

Governed by restatement.md and §6 of the licence. The part that belongs here, because it is the part people assume wrongly:

• A restatement corrects the record. It does not undo a payment. W.E.T. settles nothing, holds nothing and clears nothing. Where a contract has already settled against a value later restated, the consequence is between its counterparties.
• On-chain settlement is irreversible in a way a chart is not, and this document will not imply otherwise. A corrective record can be published; a transfer cannot be recalled.
• Corrections beyond 90 days are disclosed and analysed but not restated. Beyond that horizon the value has been relied upon and stability is worth more than the correction. A licensee gets faster notice inside that window, not a longer one.
• A restatement is an append. The original row stays, flagged superseded, and the corrective row chains off it — so a date can carry more than one row and a reference keyed on date is ambiguous exactly when it matters most. Key on checksum.

6 · Cessation, and the window nobody plans for

A cessation notice does not suspend the benchmark. For the full notice period it keeps publishing, normally, and every value published in that window is a real settlement value. A contract settling on day 12 of a 30-day notice settles against a genuine number.

That window is the part to plan for, and it is where a contract is most likely to behave in a way nobody intended — because attention is on the ending while the obligations are still running.

The procedure:

1. Notice, published, naming the reason and the last valid settlement date.
2. Direct notice to every settlement-reference licensee within one working day. A licensee does not learn this from a web page.
3. Publication continues for the full 30 days, unless the benchmark becomes uncomputable, in which case those days are refusals and §4 governs.
4. The final settlement date is named in the notice and does not move, in either direction. It is not extended to accommodate a licensee, and not brought forward to end an obligation.
5. A successor, if any, is named — with the explicit statement that the series are NOT continuous across the change. A successor is a different benchmark with a different history. A contract may not roll to it automatically; that is a decision its counterparties make.
6. Where no successor exists, W.E.T. names none. We will not nominate a third party's index as a substitute for ours. A recommendation would carry an implication of equivalence we cannot support and did not compute.
7. The full historical ledger remains published permanently, checksum chain intact — including the refusals, which are part of the record.

And the honest limit. Thirty days is the notice a publisher can give. It is not a claim that thirty days is enough for every contract, and a licensee with a longer-dated exposure should say so at licensing rather than at cessation — which is one reason §3(5) of the licence asks what depends on the number.

7 · What this document does not do

• It does not create a duty to keep computing a benchmark that cannot be computed.
• It does not guarantee a value on any given day. No such guarantee exists at any tier, and the measured availability is published rather than promised — see the availability section of data-licence.md, where the fourteen-close record is stated as WETGRI 100%, WETFED 100%, WETFRAG 85.7%, WETX 64.3%.
• It does not make W.E.T. a counterparty, a calculation agent for anyone else's contract, or a clearer of anything.
• It does not apply to unlicensed references. Those have none of the entitlements above — see §2 of settlement-reference.md.