Crude Palm Oil & Palm Kernel Oil Fractionation

Correction note — verbatim, with source locators

KBLI 2025 code 10433 'Industri Pemisahan/Fraksinasi Minyak Mentah Kelapa Sawit dan Minyak Mentah Inti Kelapa Sawit' (CPO/PKO crude-oil fractionation industry) carries canonical pp28_sources=['10433','10490'], asserting BOTH KBLI-2020 codes 10433 and 10490 as its licensing basis. This code was sampled as a Lot 2 cross-family POSITIVE ('clean') gold control (Appendix A) and DISQUALIFIED: the conductor image-verified BPS Vol.2 Lampiran 10 printed p.326 (render lampiran10_p340-340.png, Lot 2 conductor gate report Appendix A, 2026-07-18) BY EYE: '10419 (2025) <- 10490 (2020)' and '10433 (2025) <- 10433 (2020)' are SEPARATE rows — 10490 is the SOLE KBLI-2020 ancestor of a DIFFERENT 2025 code, 10419 'Industri Minyak Mentah dan Lemak Nabati dan Hewani Bukan Kelapa dan Kelapa Sawit Lainnya' (independently verified live in this session: 10419's own canonical kbli_2020_source='10490', pp28_sources=['10490']), not a co-parent of 10433. The canonical pp28_sources=['10433','10490'] on 10433 therefore associates 10490 with the WRONG child code — the THIRD contaminated 'clean' control found that session (after 52101 and 46100), all on the crosswalk-metadata layer. Its per_skala (4 rows) is HEALTHY OSS-native licensing, unaffected by this crosswalk-pointer defect, and is NOT detached by this cure ('action':'metadata_only' — per_skala stays completely untouched regardless of its non-empty state). pp28_sources corrected to ['10433'] — 10490 REMOVED (it is recorded here as belonging to 10419, never re-added to 10433). status_mapping ('MATCH_CON_AGGREGAZIONE') and intel_2026.whatChanged are left UNCHANGED by this cure: a single-element pp28_sources under a 'MATCH_CON_AGGREGAZIONE' status_mapping is an already-precedented dataset convention (47771, 52101, 60101, 60103 all carry the same shape — pp28_sources tracks the PP28-licensing-basis pointer only, separate from the full BPS crosswalk-ancestor narrative, which lives in prose), so no inconsistency is introduced; whether the merge-aware status_mapping label itself needs re-adjudication (given the 'aggregation' this code originally cited turns out, per this same finding, to have been built on a false co-parent) is flagged here as a known follow-up, not silently left undocumented — the scope of this cure is the pp28_sources removal only, per the task that authored it. This is the FIRST use of the pp28_sources_correction compiler capability, added in this same PR (scripts/kbli_filiera/cure_canonical_collisions.py), mirroring the status_mapping_correction/whatChanged_correction pattern shipped for 47771 in PR #2761. Lot 2 conductor gate (second signing + Appendix A), 2026-07-18. BPS Vol.2 sha 29f17b3b. RESIDUAL FIX (2026-07-19, M5 lane, field-vs-note contradiction closure, conductor comment on PR #2777): the 'known follow-up' flagged above ('whether the merge-aware status_mapping label itself needs re-adjudication') is now resolved — status_mapping corrected from 'MATCH_CON_AGGREGAZIONE' to 'MATCH_LANGSUNG' and whatChanged corrected to a clean 1:1 narrative, since 10490 (the co-parent this code's aggregation status depended on) belongs to 10419, not 10433: 2025-10433 is a genuine 1:1 continuation of 2020-10433 alone. pp28_sources is UNCHANGED by this residual fix (already correct at ['10433'] from #2777). aggregation_note (currently 'Dati da 10433 + 1 codici figli PP28', now stale — it describes a 2-source merge that no longer exists) is intentionally left untouched by this residual fix — cure_canonical_collisions.py's metadata_only action does not write that field; flagged here, not silently left undocumented, as a possible follow-up for a future pass.