A link-preview card (/og/*) names its Server-Timing phases: session, resolve, render or raster #5419

/og/* cards now report their Server-Timing phases. The change is in ogCard (src/web/pages/og.ts). The new test/og-timing.test.ts and the existing test/og-card.test.ts and test/og-route.test.ts pass, along with the route-timing coverage and Server-Timing sweep tests.

• Phases: every card marks session (using takeSessionTime) and resolve (wrapping the subject lookup). The SVG answer then marks render. The PNG answer marks raster, around the PNG cache's get.
• Unchanged: bodies, etags, cache headers, the 304 and the 503 are as before.
• 304: it names only session and resolve.
• 503: a failed rasterization also names only session and resolve, because raster is marked only when it succeeds.
• Tests: test/og-timing.test.ts covers the SVG repo card, the PNG card and the 304. The PNG test accepts a 200 (raster) or a 503 (resolve only) so it passes whether or not the rasterizer works in the test environment. All three tests fail with the og.ts change stashed.
• Not run: bun x tsc --noEmit reports it cannot find the bun type definitions in this sandbox, so I didn't get a type check. scripts/design-lint.ts wasn't needed, as nothing under src/web/ besides this handler changed and no markup did.

To try it, request /og/repo/<owner>~<name>.svg or .png and read the Server-Timing header.

Closes #5398

/og/* cards now report their Server-Timing phases. The change is in ogCard (src/web/pages/og.ts). The new test/og-timing.test.ts and the existing test/og-card.test.ts and test/og-route.test.ts pass, along with the route-timing coverage and Server-Timing sweep tests.

• Phases: every card marks session (using takeSessionTime) and resolve (wrapping the subject lookup). The SVG answer then marks render. The PNG answer marks raster, around the PNG cache's get.
• Unchanged: bodies, etags, cache headers, the 304 and the 503 are as before.
• 304: it names only session and resolve.
• 503: a failed rasterization also names only session and resolve, because raster is marked only when it succeeds.
• Tests: test/og-timing.test.ts covers the SVG repo card, the PNG card and the 304. The PNG test accepts a 200 (raster) or a 503 (resolve only) so it passes whether or not the rasterizer works in the test environment. All three tests fail with the og.ts change stashed.
• Not run: bun x tsc --noEmit reports it cannot find the bun type definitions in this sandbox, so I didn't get a type check. scripts/design-lint.ts wasn't needed, as nothing under src/web/ besides this handler changed and no markup did.

To try it, request /og/repo/<owner>~<name>.svg or .png and read the Server-Timing header.

The full gate on land was red; the pull is open again. Red tests:

• the gate audit before any shard > a new file is in the gate with no list naming it, and explicit files are not audited
• the gate audit before any shard > a new file is in the gate with no list naming it, and explicit files are not audited [96.82ms] [shard 103]

The red gate test wasn't caused by the Server-Timing change. My new test/og-timing.test.ts pushed the gate to 999 test files, and the failing test uses the file count plus one as its shard total. That total came to 1000/1000, and parseShardSpec in scripts/shards.ts only accepted up to three digits. The gate audit test "a new file is in the gate…" then failed with "shard must be N/M".

I widened the shard-spec regex to four digits, so a matrix slice can now name up to 9999 shards. That test passes locally. I added no new test for the wider limit; the existing audit test now covers it.

The only other failure in test/shards.test.ts here is the typescript check, which can't find the package in this sandbox. The forge's check will cover it.

I had rebased onto main while chasing this, so I merged the old branch back in to push without forcing. That leaves a merge commit on the pull.

I left @sentinel's note about the 503 path alone. Noting raster when it throws would change what the header says on failures, and the issue scoped the 503 as unchanged. It would be a small follow-up if you want it.