Synthetic feed-disposition sequence: newer, late, repeated and conflicting updates

Declared policy and initial state

This invented supplier issues comparable nonnegative integer versions for one fixed coverage. Initial accepted state is version 41, content A. Content letters stand for equality tokens, not timestamps or actual supplier hashes. An equal version with different content is held; a legitimate same-version correction would need a separately authorised policy.

Six inputs, with expected state after each

Evaluate these inputs sequentially. Rejecting a feed must leave the accepted version and content unchanged. A conflicting duplicate inside an otherwise newer feed holds it before publication under this example's whole-feed policy.

If the matrix is wider than the box, scroll horizontally to read every column. Keyboard: focus the matrix and use Left/Right.

incoming                 | expected disposition | accepted state
42, content B            | accept               | 42, B
41, content A arrives late| stale                | 42, B
42, content B repeats    | no-op                | 42, B
42, content C            | hold-version-conflict| 42, B
unknown version, D       | hold-version         | 42, B
43, E; conflicting rows  | hold-duplicate       | 42, B

Duplicate rows are a separate input check

Two Alder/00123 rows with equal price 19.99 and stock 5 receive an explicit collapse-identical disposition in the local model. If one price is 20.49 instead, either row ordering produces hold-duplicate. A Birch row with SKU 00123 is not a duplicate of Alder's row. These rules compare only the authored business fields; real feed equivalence must name every field that matters.

• Source order is not precedence unless documented.
• The local model tests sequential decisions, not locks, transactions, retries after partial writes or concurrent workers.
• No database or supplier system was exercised, and no production acceptance follows.

Priced enquiry and non-fit

A deterministic defect in an existing local version-decision function may fit the regression-test repair from £295 after bounded synthetic reproduction and a fixed quote. One new version-disposition preview interaction in an existing web app may fit the feature-preview offer from £750 with its already available isolated private preview and other eligibility conditions. Atomic multi-row publication, new connectors, concurrent database tests and historical recovery need separate scope. Send invented input order and expected state, not private payloads or credentials. Prices are untested and payment follows agreed checks and sign-off.