/* ========================================================================== SECTION RHYTHM — the whitespace at a join between two stacked sections ========================================================================== v9b brief, Round 62 (item BA2), client: "the gaps between sections are inconsistent and several are too big — the gap above the Designers Edit heading is the size I want." THE ARITHMETIC, which is the whole bug. pb-shell gives every .section a full --section-gap of padding on BOTH sides. So an ordinary join between two plain sections is padding-bottom + padding-top = TWO gaps, while the join beside a full-bleed band is ONE, purely because the band pads zero and only the neighbour contributes. Add .section--tight on --section-gap-s and you get a third and a fourth size. Measured on this theme at 1440px before this file existed: 113.8, 92.9, 56.9 and 36px joins on the home page — four sizes, none of them chosen. This was read as a tuning problem for several rounds and "fixed" one section at a time, which is why it kept coming back: no amount of nudging one section makes a sum of two terms equal a sum of one. THE FIX, in three rules. Normalise every plain section to a full gap per side [A], then halve BOTH facing sides wherever two plain sections actually meet [B] — so that join totals exactly one --section-gap, the reference, unchanged. Every other edge is left alone: a section facing a band, a hero, a breadcrumb or the end of the flow keeps its whole gap, because there is no facing padding to pair with and the whitespace beside a band lives entirely in its neighbour. Net effect: one join size everywhere, at every width. Stating it as "halve where two meet" rather than "halve everything, then put the full gap back at the edges" is what keeps this short. The edge cases are the majority of the awkward ones — first child, last child, either side of each band, under the breadcrumb — and every one of them is handled by not being mentioned. SCOPE: GLOBAL, not front-page-only. v9b scoped this to home.css and left its own inner-page sheets on the old arithmetic, because the mockup has one CSS file per page. This theme has no per-page sheets — the primitive is shared, so the same double gap lands on every template, and measurably did: the Designers' Edit page had the identical 113.8px joins at 1440. Scoping to the home page here is also not really available: this install has no static front page (show_on_front is "posts"), so front-page.php never executes and the `home` body class lands on the blog index, not on the home preview. A page-id selector would key the site's vertical rhythm to a database row. To narrow this later, put a body-level selector in front of the three rules; nothing else needs to change. SPECIFICITY, and the trap in it — v9b lost a round to this one. The weights are deliberate and are the first thing to check before editing: [A] normalise main {plain} (0,2,1) [B] halve main {plain} meeting {plain} (0,4,1) [A] must clear pb-shell's `.section` and `.section--tight`, both (0,1,0), and pb-range's `.pb-range .home-faq` at (0,2,0). The leading `main` type selector buys that on SPECIFICITY rather than on load order — this project has been bitten by order-dependent ties before, and the block sheets load BEFORE the theme parts here, not after, so "later in the cascade" is not a safe instinct. [B] must clear [A], which it does by carrying the plain-section compound twice. v9b's first cut inverted this: it wrote the BASE with :not(), did not account for :not() contributing the specificity of its argument, and the base came out heavier than the edge rules — every join silently came out half size, the reference included. WHAT COUNTS AS A PLAIN SECTION. .section and .section--tight are the shell primitive. .trick was named alongside them because the editorial callout from blocks/pb_ec_{slug}/ duplicated `padding-block: var(--section-gap)` without carrying .section, and the durable fix was said to be for that block to carry .section. IT DOES NOW (2026-08-18, the v9b restructure), and it also carries .section--recessed, so it has moved into the EXCLUDED list below — correct by this file's own rule, because the band paints a visible ground and its padding is band geometry rather than join whitespace. .trick is kept in the :is() list as a no-op safety net for any stored markup predating that change; if you are auditing dead selectors, this is one, and removing it changes nothing today. WHAT IS EXCLUDED, and why only these. A section is excluded when it paints its own visible ground, because then its padding is band GEOMETRY — interior space inside a coloured area — rather than join whitespace, and halving it would shrink the band rather than the gap: .section--recessed --surface-recessed is bone-deep, a visibly distinct well .cta-band near-black ground; its half gap centres the copy against the flush footer, stated by its own block sheet .shop-band pads zero so the photograph fills the band .section--inverse is deliberately NOT excluded: v9b flattened --surface-inverse to white, so it paints white on a white page ground and there is no visible edge — its padding really is join whitespace. Check the token before adding anything to this list; it is the difference between a band and an ordinary section. .shop-band has no block in this theme yet and is listed because it is the band the reference design pairs the REFERENCE join with, so a later port is already covered. ASSUMPTION worth knowing: [B] uses adjacent-sibling matching, so it needs the two sections to be true DOM neighbours. A non-rendering element emitted between them (a stray inline