Resolve author URLs the import filed under a different slug

[ssavutu]

The second half of the author work. #88 merged only its first commit — the category-archive redirects — so this carries the part that recovers real people, rebased onto main.

The defect

The import derives an author's slug from the WordPress login, and the logins don't agree with the names the site links by. Four distinct shapes are live in production:

URL that is linked	slug the CMS holds	why
ryan-keating	rkeating	abbreviated login
nayab-iqbal	by-nayab-iqbal	byline text used as the login
sanjana-bandi-2	sanjana-bandi	WordPress duplicate-account suffix
stefan-kusmirek	stefan-kusmirek-dev-thetriangle-org	email used as the login

Ryan Keating has 59 articles and his author page was unreachable from every URL that names him. Seven authors are affected in total.

Why display-name matching

I started with pattern rules — strip a by- prefix, strip a -2 suffix — and they were wrong. rkeating and the email-derived slug fit no prefix rule, and the next import will invent a fifth shape.

The display name is the one thing all four have in common: the CMS holds the right name in every case, only the slug is wrong. Matching on it covers all four without enumerating them, and it keeps working after the ETL fix (DrexelTriangle/wordpress-etl#63) changes the slugs.

src/utils/authorIndex.ts builds the index from a paged /v1/authors sweep and caches it in process, on the taxonomyStore pattern already in the repo: same TTL-plus-stale-fallback, same single-flight dedupe, same refusal to cache an empty result — an index with no authors would answer "no such person" for everyone. 875 authors, ~114 KB.

getAllAuthors returns undefined if any page fails rather than a partial list, because a page that failed to load is indistinguishable from an author who does not exist.

Cost and safety

It runs only after the author lookup has already missed, so a page that resolves pays nothing. It returns null when the CMS can't be read, leaving the 404 in place rather than inventing a 301 during a blip that then caches.

The duplicate-suffix strip is limited to one or two digits: real CMS slugs carry a numeric id suffix to break collisions (erik-heyman-meltzer-870), and a broader rule would be one bad import away from resolving a URL to a different person.

Verification

astro check: 0 errors. Dev server against the live CMS, on this branch rebased onto current main:

/author/ryan-keating         301 → /author/rkeating
/author/nayab-iqbal          301 → /author/by-nayab-iqbal
/author/sanjana-bandi-2      301 → /author/sanjana-bandi
/author/stefan-kusmirek      301 → /author/stefan-kusmirek-dev-thetriangle-org

/author/rkeating                 200   (real slug, untouched)
/author/erik-heyman-meltzer-870  200   (real numeric-suffix slug, not stripped)
/author/crossword                301 → /crossword   (#88 behaviour, still intact)
/author/news                     301 → /news
/author/beeboop                  404   (genuinely unknown)

🤖 Generated with Claude Code