Paternity leave (weeks)

An SI carry (ADR-0014): SI 2024/329 reg 9(a) substituted paragraph (1) WHOLE, so there is no value-level substitution for NF-1 to re-derive and the new paragraph carries no provision id in the amender's as-made document. Note what the amendment did and did not change: the TWO WEEKS are the 2002 Regulations' own figure and have not moved, while the 2024 instrument replaced the paragraph that states them in order to allow the weeks to be split. The record is dated from that substitution because it is the text being read; the earlier period, when the same two weeks had to be taken in one continuous block, waits for a parser that can read a block substitution. The locator needs two groups: the first phrase is word-for-word identical in regulation 9(1), the adoption limb, and only "under regulation 4" separates them. THE DATE IS A CALENDAR STEP, in one sentence: SI 2024/329 reg 5(1)(a) reaches only a child whose expected week of birth begins AFTER 6 April 2024, an expected week begins at midnight between Saturday and Sunday (2002 Regulations reg 2(1)), and 6 April 2024 was a Saturday, so the first week it can reach begins on Sunday 7 April 2024. The value's own first effect under D16(d) is earlier — the 2002 Regulations fixed two weeks for children born on or after 6 April 2003 (reg 3(1)), and the 2024 instrument re-enacted the figure verbatim while splitting how it may be taken — but the probe refuses that date on these documents: at 2003-04-06 it returns MISFILED, "the setting amendment key-6c65863689c3f4f7a409f0b80e94c141 took effect 2024-03-08, after the value's stated start (2003-04-06)". A date the machinery cannot evidence is not written, so the record carries the later one and says so. D-RD-3 (D18 §1) was built for exactly this row and does NOT rescue it, because the re-enactment was not verbatim: the 2002 Regulations as made read "An employee may choose to take either one week's leave or two consecutive weeks' leave in respect of a child under regulation 4", and SI 2024/329 replaced that with "a single period of leave of either one week or two weeks, or two non-consecutive periods of leave of a week each". Same figure, different words — so the origin cannot witness the text the probe located, and the relief is refused MISFILED, "the recorded origin does not state the value". Widening the locator to a phrase common to both documents would be widening a matcher to force a match. What would settle this row is a locator the 2002 wording also carries AND that still locates uniquely in the current text — or a second value period, once a parser can read a block substitution.