188-URL cleanup and hreflang on a Drupal 9 / Pantheon site

The problem

A subsidiary site of a global agribusiness brand ran Drupal 9 on Pantheon with 188 legacy URLs carrying a deprecated subsection prefix. They were bleeding into Search Console as duplicate content.

The same site needed hreflang across its international language variants, and a hover-preview regression was exposing those legacy URLs at the database level.

The approach

URL cleanup shipped across four phases through Pantheon Dev to Test to Live. Sitemap regeneration was blocked by a stale hook left behind by an installed vendor plugin, which was cleared with a url_generators config override.

Hreflang went in across every language variant with proper x-default fallback, then validated in GSC and submitted for re-crawl.

The database work was the deep part. Legacy URLs were embedded in serialized link-field data across three tables: field_button_link, field_learn_more_link, and field_link. Each table had its own column conventions, so the fix required direct DB migration scripts over SSH.

The outcome

All 188 legacy URL patterns were removed and validated in GSC, hreflang went live across all variants, and the database-level regression was closed.

The project cleaned up what users and crawlers could see, then went under the interface to remove the legacy URL references still embedded in serialized fields.