the duplicate key fallback claims every record surface keeps the first value while the disposition reader discards both and reports the record illegible

the duplicate key fallback claims every record surface keeps the first value while the disposition reader discards both and reports the record illegible

Found by the itd-189 delta ruthless review. The message text is PRE-EXISTING; the audit commissioned to bound it is BRANCH-INTRODUCED, and that is the part worth recording.

The fallback branch asserts that the lenient scanner "every record surface reads with" keeps the first value. That is false for the dsp store. issueschema.ParseDisposition returns DispositionRecord{ID: id} on a duplicated key, discarding BOTH values, under a comment that says so outright: a duplicated top-level key is malformed to every reader of this ledger.

So one abcd lint run emits two findings on one file that contradict each other. record_schema says the first value is kept, so the author concludes the disposition still stands on accepted. reading_outstanding says no reader can read the record and it carries no state. The second is the true one: standingDisposition routes the item to Unreadable and its answer decides nothing. dsp is armed in this repository's production record-lint config, so this is reachable here and not only in principle.

The recursion is the finding. 1d06fe10 split readerRefusesDuplicateKey out precisely because the previous claim had been carried across stores without checking each. Its audit enumerated adr, adm and srp, and stopped one store short of dsp -- committing the same error inside the fix for that error. Ninth surface this cycle for the standing class, and the first that is recursive.

Remedy, and the reviewer is explicit that the obvious one is wrong. Do NOT set readerRefusesDuplicateKey: true on dsp: that branch says the file is "skipped by every disposition surface", and a duplicated-key disposition is not skipped -- it is read, found illegible, and reported as an unreadable answer. It would be a second wrong sentence.

Narrow the false universal to what the rule's own godoc already licenses: this RULE'S scanner keeps only the first value, so a second line can silence a blocker armed on the value the first hides. The godoc was already the narrower and true form; the emitted message claimed more than its own documentation licensed.