Tie the AG page's rollup numbers to their labels the way the fleet page's now are #3045

#3038 fixed the fleet page's rollup tiles: every tile's number is now programmatically tied to the text that says what it counts, so a consumer reaching the number's node on its own gets the figure and its meaning together. The AG page's rollup has the identical defect and was deliberately left out of that PR, because the two pages' tile builders share no code path and one PR touching both would have been reviewed as two changes wearing one number.

The site

Darling/PerformanceMonitor.Darling.Service/wwwroot/js/pages/ag.js, rollup() at line 97:

const tile = (num, lbl, cls) =>
  el("div", { class: "tile " + (cls || "") }, [
    el("div", { class: "num", text: fmtInt(num) }),
    el("div", { class: "lbl", text: lbl }),
  ]);

Three tiles ride on it — Groups, Reporting servers, Views. The number and its label are sibling divs and nothing else. Reading order leaves the relationship inferable but not determinable, which is the same thing #3031 was about.

Correcting how this was scoped when #3031 closed

I described the fix as "the util.js hoist is the shape." There is no util.js hoist. rollupTextId and the wired tile closure both live in fleet.js (lines 480–525) and nothing in util.js exposes a tile or id helper. So this is not "call the existing shared helper" — the shared helper does not exist yet, and creating one is the actual decision this issue carries.

Two ways to go, and the choice is worth making deliberately rather than by whichever is quicker:

• Hoist rollupTextId into util.js and have both pages use it. It is eight lines with no page-specific knowledge — a label, a part name, and a used set — so it hoists cleanly. This keeps one de-duplication rule for both pages, which matters because the rule is subtle (see below).
• Duplicate the id derivation in ag.js. Cheaper to review in isolation, and leaves the two pages free to diverge. But it makes the de-duplication rule a thing that exists twice, and a rule that exists twice is a rule that will be fixed once.

I lean to the hoist, for the reason fleet.js already documents: the used set exists because two tiles sharing a label emit a duplicate id, and aria-describedby resolves a duplicate to the FIRST match — a silent mis-association rather than a visible break. That is the kind of hazard you want one implementation of. AG's three current labels are distinct, so the set is inert there today, which is exactly why a local copy would be written without it and be correct until someone adds a fourth tile.

What not to copy from fleet.js

The sub-line machinery. fleet.js's tile takes sub and title because the fleet rollup carries a coverage sub-line and a long coverage note; the AG rollup has neither, and six of the seven fleet tiles have no sub-line either. subId should not appear in an AG implementation just because it appears in the model. The part that transfers is the number describing itself with the label's id.

Also worth keeping: the reason the label rides in aria-describedby rather than aria-label or aria-labelledby — .num is a plain div, ARIA prohibits naming role=generic, so a name set there is invalid and may be dropped, while a description is a supported property. That reasoning is page-independent and should not be re-derived.

Verification this needs

Whatever shape it takes, the guard should fail on a tile whose number carries no description — not merely pass on the tiles that have one. A positive-only check cannot find an unremembered site, which is the whole reason this page survived #3031.