— FAQ answers, blog/article body copy, hero/lead paragraphs — stretching to the full content column (76-185 characters per line). This is the single shared-rule fix for that: cap the column, don't shrink the type. Scoped to
(the element type every listed violation named), and carved out from short UI paragraphs whose layout role is to fill a row (buttons rendered as
, address lines, breadcrumbs) — those were never over 68ch to begin with, so the :not() list only removes accidental risk, not real content. P12 round 15 fix (found by the exhaustive scanner's new measure check, see scan_P12.py): the original selector's :not() clauses gave it specificity (0,3,1), which beats any page-local single-class max-width rule (0,1,0) - e.g. aboutus.html's `.abt-hero__sub { max-width: 620px; }` and ~35 more `*__hero-sub`/`*__sub`/`.lead` style page-local hero-subtitle classes sitewide, each deliberately authored NARROWER than 68ch for its own layout. This rule was WINNING that fight and WIDENING those boxes to 68ch (780px at their ~19px font) instead of the page's own tighter value, which is backwards: --measure is meant to be an outer ceiling pages opt out of by setting their OWN narrower max-width, never a floor that overrides one. Wrapping the whole selector in :where() zeroes its specificity to (0,0,0) so it only ever applies where nothing more specific already set max-width - exactly the "opt-out to something narrower" behaviour the rule always intended, without changing anything for the plain, unclassed
elements (FAQ answers, blog copy) this rule exists to cap, since nothing else sets max-width on those. */ :root { --measure: 68ch; } :where(p:not(.btn):not(.breadcrumb-item):not(.office-addr)) { max-width: var(--measure); } /* A capped paragraph that was centred (hero/lead copy, CTA subtext) must stay centred in its column instead of collapsing to the left edge of its new, narrower box. */ p.text-center, .text-center > p, .lead.text-center { margin-left: auto; margin-right: auto; } /* Round 20 — the :not() clauses above were moved INSIDE :where() (they used to sit outside it), which finally gives this rule the true (0,0,0) specificity round 15's own comment always claimed it had. That correctly lets page-local narrower max-widths win (aboutus.html's .abt-hero__sub, etc.) - but it also means a page-local rule that is WIDER than --measure now wins too, which round 15's bug had been accidentally masking. Found one real instance by re-running the exhaustive scanner after this fix: the "rs-nextstep" reusable next-step block's `.rs-ns-sub` line (`.rs-nextstep .rs-ns-sub{...max-width:640px}`, page-local, identical copy-pasted rule on 16 pages) renders ~74 chars/line at 1280px on net-metering-mp - over the 68ch cap - because 640px was never checked against --measure when the component was authored. Restated here with higher specificity (two classes, matching the page-local selector's own 0,2,0) plus !important so it wins over the page-local 640px regardless of which of the 16 pages it renders on, without editing 16 separate page-local