Asymmetric clusters: when locale B exists but A doesn't, and tags assume it does

Asymmetric clusters: when locale B exists but A doesn't, and tags assume it does

A data-integrity failure that surfaces as mysterious return-tag errors: clusters where coverage is uneven across languages.

The situation. Your French and German catalogs are complete, but the Japanese one covers only a subset of products. A naive template generates the same hreflang block on every page, so French and German product pages confidently reference a ja alternate that was never published. The crawler hits a 404 (or a soft-404 fallback) where a return tag should be, and reciprocity fails for the whole cluster on that product.

The failure variants:

— Hardcoded full-locale blocks applied to pages whose translations are incomplete
— A translated page exists but at a different slug than the template assumes, so the reference 404s
— A locale's page was unpublished/expired but still referenced by its siblings

The fix is to make the alternate set dynamic and verified:

— Build each page's hreflang block from the actual set of published, indexable translations of that specific entity — not from a global locale list
— On unpublish, regenerate every sibling's block to drop the dead reference
— Validate that each referenced URL returns 200 before emitting it

Methodology note: the cheapest detection is a crawl that follows hreflang hrefs and flags any non-200 target. That single check catches the majority of asymmetric-cluster breakage we encounter.

Caveat: it's perfectly valid for clusters to differ in size page-to-page — not every product needs every language. The error isn't having a small cluster; it's claiming members that don't exist. Smaller-but-true beats complete-but-fictional.